Knowledge Graphs
Knowledge Graphs
KNOWLEDGE
Table of Contents
GRAPHS
Aidan
Aidan Hogan
Hogan || Eva
Eva Blomqvist
Blomqvist || Michael
Michael Cochez
Cochez
Claudia
Claudia d’Amato | Gerard de Melo | Claudio Gutierrez
d’Amato | Gerard de Melo | Claudio Gutierrez
Sabrina
Sabrina Kirrane
Kirrane || José
José Emilio
Emilio Labra
Labra Gayo
Gayo || Roberto
Roberto Navigli
Navigli
Sebastian
Sebastian Neumaier
Neumaier || Axel-Cyrille
Axel-Cyrille Ngonga
Ngonga Ngomo
Ngomo
Axel
Axel Polleres
Polleres || Sabbir
Sabbir M.
M. Rashid
Rashid || Anisa
Anisa Rula
Rula
Lukas
Lukas Schmelzeisen
Schmelzeisen || Juan
Juan Sequeda
Sequeda
Steffen
Steffen Staab
Staab || Antoine
Antoine Zimmermann
Zimmermann
https://kgbook.org
K nowledge
Give us feedback!
1/150
G
2/15/23, 5:13 PM Knowledge Graphs
raphs
Aidan Hogan
IMFD, DCC, Universidad de Chile
Eva Blomqvist
Linköping University
Michael Cochez
Vrije Universiteit Amsterdam and Discovery Lab, Elsevier
Table of Contents
Claudia d’Amato
University of Bari
Gerard de Melo
HPI, University of Potsdam and Rutgers University
Claudio Gutierrez
IMFD, DCC, Universidad de Chile
Sabrina Kirrane
WU Vienna
Roberto Navigli
Sapienza University of Rome
Sebastian Neumaier
St. Pölten University of Applied Sciences
Axel Polleres
WU Vienna
Sabbir M. Rashid
Tetherless World Constellation, Rensselaer Polytechnic Institute
Anisa Rula
University of Brescia
Lukas Schmelzeisen
Universität Stuttgart
Juan Sequeda
data.world
Steffen Staab
Universität Stuttgart and University of Southampton
Antoine Zimmermann
École des mines de Saint-Étienne
Give us feedback!
https://kgbook.org 2/150
2/15/23, 5:13 PM Knowledge Graphs
Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia d’Amato, Gerard de Melo, Claudio
Gutierrez, Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier,
Axel-Cyrille Ngonga Ngomo, Axel Polleres, Sabbir M. Rashid, Anisa Rula, Lukas
Schmelzeisen, Juan Sequeda, Steffen Staab, Antoine Zimmermann (2021) Knowledge
Graphs, Synthesis Lectures on Data, Semantics, and Knowledge, No. 22, 1–237, DOI:
10.2200/S01125ED1V01Y202109DSK022, Morgan & Claypool
@book{kg-book,
author = {Hogan, Aidan and Blomqvist, Eva and Cochez, Michael and
d'Amato, Claudia and de Melo, Gerard and Guti\'errez, Claudio and
Table of Contents
Kirrane, Sabrina and Labra Gayo, Jos\'e Emilio and Navigli, Roberto and
Neumaier, Sebastian and Ngonga Ngomo, Axel-Cyrille and Polleres, Axel and
Rashid, Sabbir M. and Rula, Anisa and Schmelzeisen, Lukas and
Sequeda, Juan F. and Staab, Steffen and Zimmermann, Antoine},
doi = {10.2200/S01125ED1V01Y202109DSK022},
isbn = {9781636392363},
language = {English},
number = {22},
numpages = {237},
publisher = {Morgan \& Claypool},
series = {Synthesis Lectures on Data, Semantics, and Knowledge},
title = {{K}nowledge {G}raphs},
url = {https://kgbook.org/},
year = {2021}
}
ISBN paperback:
9781636392356
ISBN ebook:
9781636392363
ISBN hardcover:
9781636392370
Access options
HTML version:
You are currently reading the free HTML version of the book, the most recent of which is avail‐
able at https://kgbook.org/.* note *
You can see the scripts that generate this
PDF Version: comments as new issues. You can also address y
tutorial [at] googlegroups [dot] com
You can download or buy the book from Morgan & Claypool. Academiccan
andbeCorporate
found onlicences
Github as well.
are available.
Hard copy:
You can order from Morgan & Claypool or Amazon.
Give us feedback!
☙ Abstract ❧
https://kgbook.org 3/150
2/15/23, 5:13 PM Knowledge Graphs
Abstract ❧
This book provides a comprehensive and accessible introduction to knowledge graphs, which have
recently garnered notable attention from both industry and academia.
Knowledge graphs are founded on the principle of applying a graph-based abstraction to data, and
are now broadly deployed in scenarios that require integrating and extracting value from multiple,
diverse sources of data at large scale. The book defines knowledge graphs and provides a high-level
overview of how they are used. It presents and contrasts popular graph models that are commonly
used to represent data as graphs, and the languages by which they can be queried before describing
how the resulting data graph can be enhanced with notions of schema, identity, and context. The
book discusses how ontologies and rules can be used to encode knowledge as well as how inductive
techniques — based on statistics, graph analytics, machine learning, etc. — can be used to encode
and extract knowledge. It covers techniques for the creation, enrichment, assessment, and refinement
of knowledge graphs and surveys recent open and enterprise knowledge graphs and the industries or
applications within which they have been most widely adopted. The book closes by discussing the
current limitations and future directions along which knowledge graphs are likely to evolve.
This book is aimed at students, researchers, and practitioners who wish to learn more about knowl‐
Table of Contents
edge graphs and how they facilitate extracting value from diverse data at large scale. To make the
book accessible for newcomers, running examples and graphical notation are used throughout.
Formal definitions and extensive references are also provided for those who opt to delve more deeply
into specific topics.
☙ Keywords ❧
knowledge graphs, graph databases, knowledge graph embeddings, graph neural networks, ontolo‐
gies, knowledge graph refinement, knowledge graph quality, knowledge bases, artificial intelligence,
semantic web, machine learning
Preface
The origins of this book can be traced back to a Dagstuhl Seminar, held in 2018, on the topic of
Knowledge Graphs. At the time of the seminar, the topic was quickly becoming mainstream in acade‐
mia and industry, but there were conflicting messages as to what a “knowledge graph” was. Much of
the discussion of the seminar centred on this question, and there were divergent opinions as to how
knowledge graphs could (or should) be defined; how they relate to previous concepts such as graph
databases, knowledge bases, ontologies, RDF graphs, property graphs, semantic networks, etc.; and
how the emerging area of Knowledge Graphs should be positioned with respect to the established ar‐
eas of Artificial Intelligence, Big Data, Databases, Graph Theory, Logic, Machine Learning,
Knowledge Representation, Natural Language Processing, Networks (in their various forms), and the
Semantic Web. As the discussion continued, a consensus began to emerge: Knowledge Graphs, as a
topic, involves a novel confluence of techniques stemming from previously disparate scientific com‐
munities, with the unifying goal of developing novel graph-based techniques for better integrating
and extracting value from diverse knowledge sources at large scale.
As a follow-up to the seminar, the attendees agreed that in order to foster this unifying view of
Knowledge Graphs, there was a need for a manuscript that would serve as a general introduction to
the area. This manuscript would:
Give us feedback!
https://kgbook.org 4/150
2/15/23, 5:13 PM Knowledge Graphs
motivate knowledge graphs and the value of abstracting data as graphs;
survey the historical context of knowledge graphs and the key initiatives leading to their
popularisation;
draw together disparate views of knowledge graphs into a unifying definition;
provide an introduction to the key techniques that knowledge graphs enable, relating to query‐
ing, validation, reasoning, learning, refinement, enrichment, quality assessment, and more
besides;
describe how knowledge graphs are used in practice, surveying the companies using knowledge
graphs, the applications they are used for, the open knowledge graphs that have been pub‐
lished, etc.;
delineate future research directions for knowledge graphs.
The manuscript would then serve as an introductory text for students, practitioners and researchers
new to the area, helping to form a consensus in terms of what is a knowledge graph, laying the foun‐
dations for future developments.
The goal of preparing this manuscript was an ambitious one, and involved drawing together and dis‐
Table of Contents
tilling down a vast amount of literature on a diverse range of topics into a set of key concepts de‐
scribed in an accessible way. For this reason, the manuscript has been prepared by many authors,
who have lent their knowledge and expertise to the preparation of specific sections. A short version
of the manuscript was first published as a tutorial paper [Hogan et al., 2021], consisting of an
abridged version of the first five chapters of this book, along with a summary of how knowledge
graphs are used in practice, and conclusions. However, there was not enough space to describe all of
the important developments in the area. This led us to publish this book, which further includes top‐
ics relating to the creation, enrichment, quality assessment, refinement and publication of knowledge
graphs, as well as formal definitions, a historical perspective, and extended discussion throughout.
The book is divided into ten chapters. The first chapter provides a general introduction to the area,
defines the concept of a “knowledge graph”, and provides a high-level overview of how knowledge
graphs are currently being used. The second chapter presents and contrasts popular graph models
that are commonly used to represent data as graphs, and the languages by which they can be
queried. The third chapter describes how the resulting data graph can be enhanced with notions of
schema, identity and context. The fourth chapter discusses how ontologies and rules can be used to
encode knowledge, and how they enable deductive forms of reasoning. The fifth chapter delves into
how inductive techniques – based on statistics, graph analytics, machine learning, etc. – can be used
to encode and extract knowledge. The sixth chapter is dedicated to techniques for the creation and
enrichment of knowledge graphs from legacy sources of data. The seventh chapter enumerates a va‐
riety of quality measures that can be used to assess a knowledge graph in terms of its fitness for use
in a variety of applications. The eighth chapter presents key methods for the refinement of knowl‐
edge graphs, with the goal of improving their completeness and correctness. The ninth chapter pro‐
vides a survey of the open and enterprise knowledge graphs that have emerged in recent years, along
with the industries within which, and the applications for which, they have been most widely adopt‐
ed. The tenth chapter wraps up the book with discussion of the current limitations and future direc‐
tions along which knowledge graphs are likely to evolve. An appendix further covers knowledge
graphs from an historical perspective, establishing their significance in the broader context of the
academic study of data and knowledge, as well as surveying prior definitions of “knowledge graphs”
from the literature.
A key aim of this book is to be accessible to a broader audience. While background knowledge of re‐
lated topics such as Databases, Logic, Machine Learning, Semantic Web, etc., will help to understand
some of the particular topics mentioned, such a background is not necessary to follow the general
concepts described within. The book aims to motivate and illustrate the various concepts it intro‐
duces from a practical perspective, and in order to be as accessible as possible, relies heavily on an
example-driven presentation using a graphical notation. For the reader wishing to dig more into the
technical minutiae, we complement this discussion with formal definitions throughout; however, the
reader more interested in understanding the general concepts and their rationale will find the discus‐
sion to be self-contained if they choose to skip the definitions presented in visually distinctive boxes.
Give us feedback!
https://kgbook.org 5/150
2/15/23, 5:13 PM Knowledge Graphs
The book serves as an entry point for those new to the topic, and may thus serve as a useful textbook
for university courses, for researchers who are venturing into the topic for the first time, and for
practitioners who wish to understand more about how knowledge graphs might be of use within
their company or organisation, or indeed, how to maximise the value of the knowledge graphs that
they are currently developing. Readers who are already active within specific sub-areas of
Knowledge Graphs may further appreciate the technical definitions included, the references to other
literature provided, and the broader perspective that this book offers in terms of the other related
sub-areas and how they complement each other.
By drawing together diverse techniques from disparate areas, Knowledge Graphs has become an ex‐
citing topic in terms of both research and applications. We expect to see growing interest on this top‐
ic as the years advance, and indeed hope that this book will help to more firmly establish the foun‐
dations of this topic, and to foster future developments upon these foundations, potentially by its
readers.
Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia d’Amato, Gerard de Melo, Claudio Gutierrez,
Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier, Axel-
Table of Contents
Cyrille Ngonga Ngomo, Axel Polleres, Sabbir M. Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequeda,
Steffen Staab, Antoine Zimmermann
September 2021
Acknowledgements
We thank the organisers and attendees of the Dagstuhl Seminar on “Knowledge Graphs”. We also
thank those who provided feedback on this content.
Hogan was funded by Fondecyt Grant No. 1181896. Hogan & Gutierrez were funded by ANID –
Millennium Science Initiative Program – Code ICN17_002. Cochez did part of the work while em‐
ployed at Fraunhofer FIT, Germany and was later partially funded by Elsevier’s Discovery Lab.
Kirrane, Ngonga Ngomo, Polleres & Staab received funding through the project “KnowGraphs” from
the European Union’s Horizon programme under the Marie Skłodowska-Curie grant agreement
No. 860801. Kirrane & Polleres were supported by the European Union’s Horizon 2020 research and
innovation programme under grant 731601. Labra was supported by the Spanish Ministry of
Economy and Competitiveness (Society challenges: TIN2017-88877-R). Navigli was supported by the
MOUSSE ERC Grant No. 726487 under the European Union’s Horizon 2020 research and innovation
programme. Rashid was supported by IBM Research AI through the AI Horizons Network.
Schmelzeisen was supported by the German Research Foundation (DFG) grant STA 572/18-1.
Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia d’Amato, Gerard de Melo, Claudio Gutierrez,
Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier, Axel-
Cyrille Ngonga Ngomo, Axel Polleres, Sabbir M. Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequeda,
Steffen Staab, Antoine Zimmermann
September 2021
https://kgbook.org 6/150
2/15/23, 5:13 PM Knowledge Graphs
Introduction
Though the phrase “knowledge graph” has been used in the literature since at least 1972 [Schneider,
1973], the modern incarnation of the phrase stems from the 2012 announcement of the Google
Knowledge Graph [Singhal, 2012], followed by further announcements of knowledge graphs by
Airbnb [Chang, 2018], Amazon [Krishnan, 2018], eBay [Pittman et al., 2017], Facebook [Noy et al.,
2019], IBM [Devarajan, 2017], LinkedIn [He et al., 2016], Microsoft [Shrivastava, 2017],
Uber [Hamad et al., 2018], and more besides. The growing industrial uptake of the concept proved
difficult for academia to ignore: more and more scientific literature is being published on knowledge
graphs, which includes books (e.g., [Pan et al., 2017, Qi et al., 2021, Fensel et al., 2020, Kejriwal et
al., 2021]), as well as papers outlining definitions (e.g., [Ehrlinger and Wöß, 2016]), novel tech‐
niques (e.g., [Pujara et al., 2013, Wang et al., 2014, Lin et al., 2015]), and surveys of specific aspects
of knowledge graphs (e.g., [Paulheim, 2017, Wang et al., 2017]).
Underlying all such developments is the core idea of using graphs to represent data, often enhanced
with some way to explicitly represent knowledge [Noy et al., 2019]. The result is most often used in
Table of Contents
application scenarios that involve integrating, managing and extracting value from diverse sources of
data at large scale [Noy et al., 2019]. Employing a graph-based abstraction of knowledge has numer‐
ous benefits in such settings when compared with, for example, a relational model or NoSQL alterna‐
tives. Graphs provide a concise and intuitive abstraction for a variety of domains, where edges cap‐
ture the (potentially cyclical) relations between the entities inherent in social data, biological inter‐
actions, bibliographical citations and co-authorships, transport networks, and so forth [Angles and
Gutierrez, 2008]. Graphs allow maintainers to postpone the definition of a schema, allowing the data
– and its scope – to evolve in a more flexible manner than typically possible in a relational setting,
particularly for capturing incomplete knowledge [Abiteboul, 1997]. Unlike (other) NoSQL models,
specialised graph query languages support not only standard relational operators (joins, unions, pro‐
jections, etc.), but also navigational operators for recursively finding entities connected through arbi‐
trary-length paths [Angles et al., 2017]. Standard knowledge representation formalisms – such as on‐
tologies [Hitzler et al., 2012, Brickley and Guha, 2014, Mungall et al., 2012] and rules [Horrocks et
al., 2004, Kifer and Boley, 2013] – can be employed to define and reason about the semantics of the
terms used to label and describe the nodes and edges in the graph. Scalable frameworks for graph
analytics [Malewicz et al., 2010, Xin et al., 2013a, Stutz et al., 2016] can be leveraged for computing
centrality, clustering, summarisation, etc., in order to gain insights about the domain being de‐
scribed. Various representations have also been developed that support applying machine learning
techniques both directly and indirectly over graphs [Wang et al., 2017, Wu et al., 2019].
In summary, the decision to build and use a knowledge graph opens up a range of techniques that
can be brought to bear for integrating and extracting value from diverse sources of data at large
scale. The goal of this book is to motivate and give a comprehensive introduction to knowledge
graphs: to describe their foundational data models and how they can be queried; to discuss represen‐
tations relating to schema, identity, and context; to discuss deductive and inductive ways to make
knowledge explicit; to present a variety of techniques that can be used for the creation and enrich‐
ment of graph-structured data; to describe how the quality of knowledge graphs can be discerned
and how they can be refined; to discuss standards and best practices by which knowledge graphs can
be published; and to provide an overview of existing knowledge graphs found in practice. Our in‐
tended audience includes researchers and practitioners who are new to knowledge graphs. As such,
we do not assume that readers have specific expertise on knowledge graphs.
Knowledge graph. The definition of a “knowledge graph” remains contentious [Ehrlinger and Wöß,
2016, Bonatti et al., 2018, Bergman, 2019], where a number of (sometimes conflicting) definitions
have emerged, varying from specific technical proposals to more inclusive general proposals; we ad‐
dress these prior definitions in Appendix A. Herein we adopt an inclusive definition, where we view
a knowledge graph as a graph of data intended to accumulate and convey knowledge of the real world,
whose nodes represent entities of interest and whose edges represent relations between these entities. The
graph of data (aka data graph) conforms to a graph-based data model, which may be a directed edge-
labelled graph, a property graph, etc. (we discuss concrete alternatives in Chapter 2). By knowledge, we
Give
refer to something that is known. Such knowledge may be accumulated from external sources, or ex‐ us feedback!
https://kgbook.org 7/150
2/15/23, 5:13 PM Knowledge Graphs
tracted from the knowledge graph itself. Knowledge may be composed of simple statements, such as
“Santiago is the capital of Chile”, or quantified statements, such as “all capitals are cities”. Simple state‐
ments can be accumulated as edges in the data graph. If the knowledge graph intends to accumulate
quantified statements, a more expressive way to represent knowledge – such as ontologies or rules – is
required. Deductive methods can then be used to entail and accumulate further knowledge (e.g.,
“Santiago is a city”). Additional knowledge – based on simple or quantified statements – can also be
extracted from and accumulated by the knowledge graph using inductive methods.
Knowledge graphs are often assembled from numerous sources, and as a result, can be highly diverse
in terms of structure and granularity. To address this diversity, representations of schema, identity,
and context often play a key role, where a schema defines a high-level structure for the knowledge
graph, identity denotes which nodes in the graph (or in external sources) refer to the same real-world
entity, while context may indicate a specific setting in which some unit of knowledge is held true. As
aforementioned, effective methods for extraction, enrichment, quality assessment, and refinement are re‐
quired for a knowledge graph to grow and improve over time.
In practice. Knowledge graphs aim to serve as an ever-evolving shared substrate of knowledge with‐
Table of Contents
in an organisation or community [Noy et al., 2019]. We distinguish two types of knowledge graphs
in practice: open knowledge graphs and enterprise knowledge graphs. Open knowledge graphs are pub‐
lished online, making their content accessible for the public good. The most prominent examples –
DBpedia [Lehmann et al., 2015], Freebase [Bollacker et al., 2007b], Wikidata [Vrandečić and
Krötzsch, 2014], YAGO [Hoffart et al., 2011], etc. – cover many domains and are either extracted
from Wikipedia [Lehmann et al., 2015, Hoffart et al., 2011], or built by communities of
volunteers [Bollacker et al., 2007b, Vrandečić and Krötzsch, 2014]. Open knowledge graphs have
also been published within specific domains, such as media [Raimond et al., 2014],
government [Hendler et al., 2012, Shadbolt and O'Hara, 2013], geography [Stadler et al., 2012],
tourism [Lu et al., 2016, Kärle et al., 2018, Maturana et al., 2018, Zhang et al., 2019], life
sciences [Callahan et al., 2013], and more besides. Enterprise knowledge graphs are typically inter‐
nal to a company and applied for commercial use-cases [Noy et al., 2019]. Prominent industries us‐
ing enterprise knowledge graphs include Web search (e.g., Bing [Shrivastava, 2017],
Google [Singhal, 2012]), commerce (e.g., Airbnb [Chang, 2018], Amazon [Krishnan, 2018, Dong,
2019], eBay [Pittman et al., 2017], Uber [Hamad et al., 2018]), social networks (e.g.,
Facebook [Noy et al., 2019], LinkedIn [He et al., 2016]), finance (e.g., Accenture [Okorafor and Ray,
2019], Banca d’Italia [Bellomarini et al., 2019], Bloomberg [Meij, 2019], Capital One [Branum and
Sehon, 2019], Wells Fargo [Newman, 2019]), among others. Applications include
search [Shrivastava, 2017, Singhal, 2012], recommendations [Chang, 2018, Hamad et al., 2018, He
et al., 2016, Noy et al., 2019], personal agents [Pittman et al., 2017], advertising [He et al., 2016],
business analytics [He et al., 2016], risk assessment [Tobin, 2017, Dalgliesh, 2016],
automation [Henson et al., 2019], and more besides. We will provide more details on the use of
knowledge graphs in practice in Chapter 10.
Running example. To keep the discussion accessible, throughout the book, we present concrete ex‐
amples in the context of a hypothetical knowledge graph relating to tourism in Chile (loosely in‐
spired by related use-cases [Kärle et al., 2018, Lu et al., 2016]). The knowledge graph is managed by
a tourism board that aims to increase tourism in the country and promote new attractions in strate‐
gic areas. The knowledge graph itself will eventually describe tourist attractions, cultural events, ser‐
vices, businesses, travel routes, etc. Some applications the organisation envisages are to:
create a tourism portal that allows visitors to search for attractions, upcoming events, and oth‐
er related services (in multiple languages);
gain insights into tourism demographics in terms of season, nationalities, etc.;
analyse sentiment about tourist attractions, including positive reviews, summaries of com‐
plaints about events and services, crime reports, etc.;
understand tourism trajectories: the sequence of attractions, events, etc., that tourists often
visit;
cross-reference these tourism trajectories with currently available flights, buses, etc., to suggest
new strategic routes for public transport;
Give us feedback!
offer personalised recommendations of places to visit;
https://kgbook.org 8/150
2/15/23, 5:13 PM Knowledge Graphs
and so forth.
Chapter 2: outlines graph data models and the languages used to query them.
Chapter 3: describes representations of schema, identity, and context for graphs.
Chapter 4: presents deductive formalisms for representing and entailing knowledge.
Chapter 5: describes inductive techniques for learning from graphs.
Chapter 6: discusses the creation and enrichment of knowledge graphs.
Chapter 7: enumerates dimensions for assessing knowledge graph quality.
Chapter 8: discusses various techniques for knowledge graph refinement.
Chapter 9: introduces principles and protocols for publishing knowledge graphs.
Chapter 10: surveys some prominent knowledge graphs and their applications.
Chapter 11: concludes with future directions for knowledge graphs.
Appendix A: outlines the historical background for knowledge graphs.
Table of Contents
Chapter 2
Data Graphs
At the foundation of any knowledge graph is the principle of first applying a graph abstraction to
data, resulting in an initial data graph. We now discuss a selection of graph-structured data models
that are commonly used in practice to represent data graphs. We then discuss the primitives that
form the basis of graph query languages used to interrogate such data graphs.
2.1 Models
Leaving aside graphs, let us assume that the tourism board from our running example has not yet de‐
cided how to model relevant data about attractions, events, services, etc. The board first considers
using a tabular structure – in particular, relational databases – to represent the required data, and
though they do not know precisely what data they will need to capture, they begin to design an ini‐
tial relational schema. They begin with an Event table with five columns:
where name and start together form the primary key of the table in order to uniquely identify recur‐
ring events. But as they start to populate the data, they encounter various issues: events may have
multiple names (e.g., in different languages), events may have multiple venues, they may not yet
know the start and end date-times for future events, events may have multiple types, and so forth.
Incrementally addressing these modelling issues as the data become more diverse, they generate in‐
ternal identifiers for events and adapt their relational schema until they have:
With the above schema, the organisation can now model events with 0−n names, venues, and types,
and 0−1 start dates and end dates (without needing relational nulls).
Along the way, the board has to incrementally change the schema several times in order to support
new sources of data. Each such change requires a costly remodelling, reloading, and reindexing of
data; here we only considered one table. The tourism board struggles with the relational model be‐
cause they do not know, a priori, what data will need to be modelled or what sources they will use.
But once they reach the latter relational schema, the board finds that they can integrate further
Give us feedback!
https://kgbook.org 9/150
2/15/23, 5:13 PM Knowledge Graphs
sources without more changes: with minimal assumptions on multiplicities (1−1, 1−n, etc.) this
schema offers a lot of flexibility for integrating incomplete and diverse data.
In fact, the refined, flexible schema that the board ends up with – as shown in (2.1) – is modelling a
set of binary relations between entities, which indeed can be viewed as modelling a graph. By in‐
stead adopting a graph data model from the outset, the board could forgo the need for an upfront
schema, and could define any (binary) relation between any pair of entities at any time.
We now introduce graph data models popular in practice [Angles et al., 2017].
A directed edge-labelled graph (sometimes known as a multi-relational graph [Nickel and Tresp, 2013,
Bordes et al., 2013, Balazevic et al., 2019a]) is defined as a set of nodes – like Santiago , Arica ,
EID16 , 2018-03-22 12:00 – and a set of directed labelled edges between those nodes, like Santa Lucía –
city ➛ Santiago . In the case of knowledge graphs, nodes are used to represent entities and edges are
used to represent (binary) relations between those entities. Figure 2.1 provides an example of how
Table of Contents
the tourism board could model some relevant event data as a directed edge-labelled graph. The
graph includes data about the names, types, start and end date-times, and venues
note 1for events. 1 Adding
We represent bidirectional edges as Viña d
information to such a graph typically involves adding new nodes and edgesdepicts
(with some
two exceptions
directed edges: Viña del Mar –bus ➛
discussed later). Representing incomplete information requires simply omitting a particular
while edge; for
some naming conventions recommend m
verb,
example, the graph does not yet define a start/end date-time for the Food Truck such as has venue or is valid from, in this
festival.
omit the “has” and “is” verbs from such labels,
Figure 2.1: Directed edge-labelled graph describing events and their venues
Modelling data as a graph in this way offers more flexibility for integrating new sources of data,
compared to the standard relational model, where a schema must be defined upfront and followed at
each step. While other structured data models such as trees (XML, JSON, etc.) would offer similar
flexibility, graphs do not require organising the data hierarchically (should venue be a parent, child,
or sibling of type for example?). They also allow cycles to be represented and queried (e.g., note the
directed cycle in the routes between Santiago, Arica, and Viña del Mar).
A standardised data model based on directed edge-labelled graphs is the Resource Description
Framework (RDF) [Cyganiak et al., 2014], which has been recommended by the W3C. The RDF
model defines different types of nodes, including Internationalized Resource Identifiers (IRIs) [Dürst
and Suignard, 2005] which allow for global identification of entities on the Web; literals, which al‐
low for representing strings (with or without language tags) and other datatype values (integers,
dates, etc.); and blank nodes, which are anonymous nodes that are not assigned an identifier (for ex‐
ample, rather than create internal identifiers like EID15, EID16, in RDF, we have the option to use
blank nodes). We will discuss these different types of nodes further in Section 3.2 when we speak
about issues relating to identity.
We now formally define a directed edge-labelled graph, where we denote by Con a countably
infinite set of constants.
Give us feedback!
https://kgbook.org 10/150
2/15/23, 5:13 PM Knowledge Graphs
Example 2.1 — In reference to Figure 2.1, the set of nodes V has 15 elements, including Arica,
EID16, etc. The set of edges E has 23 triples, including (Arica, flight, Santiago). Bidirectional
edges are represented with two edges. The set of edge labels L has 8 elements, including start,
flight, etc.
Definition 2.1 does not state that V and L are disjoint: though not present in the example, a
node can also serve as an edge-label. The definition also permits that nodes and edge labels can
be present without any associated edge. Either restriction could be explicitly stated – if neces‐
sary – in a particular application while still conforming to a directed edge-labelled graph.
For ease of presentation, we may treat a set of (directed labelled) edges E ⊆ V × L × V as a di‐
rected edge-labelled graph (V , E, L), in which case we refer to the graph induced by E assuming
Table of Contents
that V and L contain all and only those nodes and edge labels, respectively, used in E . We may
similarly apply set operators on directed edge-labelled graphs, which should be interpreted as
applying to their sets of edges; for example, given G1 = (V1 , E1 , L1 ) and G2 = (V2 , E2 , L2 ), by
A heterogeneous graph [Hussein et al., 2018, Wang et al., 2019, Yang et al., 2020] (or heterogeneous
information network [Sun et al., 2011, Sun and Han, 2012]) is a directed graph where each node and
edge is assigned one type. Heterogeneous graphs are thus akin to directed edge-labelled graphs –
with edge labels corresponding to edge types – but where the type of node forms part of the graph
model itself, rather than being expressed with a relation (as seen in Figure 2.2). An edge is called ho‐
mogeneous if it is between two nodes of the same type (e.g., borders in Figure 2.2); otherwise it is
called heterogeneous (e.g., capital in Figure 2.2). Heterogeneous graphs allow for partitioning nodes
according to their type, for example, for the purposes of machine learning tasks [Hussein et al.,
2018, Wang et al., 2019, Yang et al., 2020]. Conversely, such graphs typically only support a one-to-
one relation between nodes and types, which is not the case for directed edge-labelled graphs (see,
for example, the node Santiago with zero types and EID15 with multiple types in Figure 2.1).
Give us feedback!
https://kgbook.org 11/150
2/15/23, 5:13 PM Knowledge Graphs
Example 2.2 — In reference to Figure 2.2b, the set of nodes V has three elements: Santiago,
Chile, and Perú. The set of edges E has 3 triples, including (Santiago, capital, Chile). The
set of edge labels L has 4 elements: capital, borders, City, Country. Finally, with respect to
the node labels, l(Santiago) = City, l(Chile) = Country, and l(Perú) = Country.
In heterogeneous graphs, edge and node labels are often called types. By rather defining edges
with labels as per directed edge-labelled graphs – rather than separately labelling edges with l –
two nodes can be related by n edges with n different labels; for example, we can represent both
(Santiago, capital, Chile) and (Santiago, country, Chile) as edges in the heteroge‐
neous graph.
Table of Contents
Property graphs constitute an alternative graph model that offers additional flexibility when mod‐
elling more complex relations. Consider integrating incoming data that provide further details on
which companies offer fares on which flights, allowing the board to better understand available
routes between cities (for example, on national airlines). In the case of directed edge-labelled graphs,
we cannot directly annotate an edge like Santiago –flight ➛ Arica with the company (or companies) of‐
fering that route. But we could add a new node denoting a flight, connect it with the source, destina‐
tion, companies, and mode, as shown in Figure 2.3a. Applying this modelling to all routes in
Figure 2.1 would, however, involve significant changes.
The property graph model was thus proposed to offer additional flexibility when modelling data as a
graph [Miller, 2013, Angles et al., 2017]. A property graph allows a set of property–value pairs and a
label to be associated with both nodes and edges. Figure 2.3b depicts an example of a property graph
with data analogous to Figure 2.3a. We use property–value pairs on edges to model the companies.
The type of relation is captured by the label flight. We further use node labels to indicate the
types of the two nodes, and property–value pairs for their latitude and longitude.
Give us feedback!
https://kgbook.org 12/150
2/15/23, 5:13 PM Knowledge Graphs
Property graphs are prominently used in graph databases, such as Neo4j [Miller, 2013, Angles et al.,
2017]. Property graphs can be converted to/from directed edge-labelled graphs [Hernández et al.,
2015, Angles et al., 2019] (per, e.g., Figure 2.3b). In summary, directed edge-labelled graphs offer a
more minimal model, while property graphs offer a more flexible one. Often the choice of model will
be secondary to other practical factors, such as the implementations available for different models,
etc.
Unlike previous definitions [Angles et al., 2017], we allow a node or edge to have several values
for a given property. In practice, systems like Neo4j [Miller, 2013] may rather support this by
allowing a single array (i.e., list) of values.
Although multiple directed edge-labelled graphs can be merged by taking their union, it is often de‐
sirable to manage several graphs rather than one monolithic graph; for example, it may be beneficial
to manage multiple graphs from different sources, making it possible to update or refine data from
one source, to distinguish untrustworthy sources from more trustworthy ones, and so forth. A graph
dataset then consists of a set of named graphs and a default graph. Each named graph is a pair of a
graph ID and a graph. The default graph is a graph without an ID, and is referenced “by default” if a
graph ID is not specified. Figure 2.4 provides an example where events and routes are stored in two
named graphs, and the default graph manages metadata about the named graphs. Graph names can
also be used as nodes in a graph. Furthermore, nodes and edges can be repeated across graphs,
where the same node in different graphs will typically refer to the same entity, allowing data on that
entity to be integrated when merging graphs. Though the example depicts a dataset of directed edge-
labelled graphs, the concept generalises straightforwardly to datasets of other types of graphs.
Give us feedback!
https://kgbook.org 13/150
2/15/23, 5:13 PM Knowledge Graphs
Table of Contents
Figure 2.4: Graph dataset based on directed edge-labelled graphs with two named graphs and a default
graph describing events and routes
An RDF dataset is a graph dataset model standardised by the W3C [Cyganiak et al., 2014] where
each graph is an RDF graph, and graph names can be blank nodes or IRIs. A prominent use-case for
RDF datasets is to manage and query Linked Data composed of interlinked documents of RDF graphs
spanning the Web. When dealing with Web data, tracking the source of data becomes of key impor‐
tance [Dividino et al., 2009, Bonatti et al., 2011, Zimmermann et al., 2012]. We will discuss Linked
Data later in Section 3.2 and further discuss provenance in Section 3.3.
We more formally define a graph dataset. We assume that all data graphs featured in a given
graph dataset follow the same model (directed edge-labelled graph, heterogeneous graph, prop‐
erty graph, etc).
the empty set, or a set of named graphs {(n1 , G1 ), … (nk , Gk )} (k > 0) such that if i =
j then
= nj (for all 1 ≤ i ≤ k , 1 ≤ j ≤ k ).
ni
Example 2.4 — Figure 2.4 provides an example of a directed edge-labelled graph dataset D consist‐
ing of two named graphs and a default graph. The default graph does not have a name associated with
it. The two graph names are Events and Routes; these are also used as nodes in the default graph.
The previous models are popular examples of graph representations. Other graph data models exist
with complex nodes that may contain individual edges [Angles and Gutierrez, 2008, Hartig and
Thompson, 2014] or nested graphs [Angles and Gutierrez, 2008, Berners-Lee and Connolly, 2011]
(sometimes called hypernodes [Levene and Poulovassilis, 1989]. Likewise the mathematical notion of
a hypergraph defines complex edges that connect sets rather than pairs of nodes. In our view, a knowl‐
edge graph can adopt any such graph data model based on nodes and edges: often data can be con‐
verted from one model to another (see Figure 2.3a vs. Figure 2.3b). In the rest of the paper, we pre‐
fer discussing directed edge-labelled graphs given their relative succinctness, but most discussion ex‐
tends naturally to other models.
https://kgbook.org 14/150
2/15/23, 5:13 PM Knowledge Graphs
A variety of techniques have been proposed for storing and indexing graphs, facilitating the efficient
evaluation of queries (as discussed next). Directed edge-labelled graphs can be stored in relational
databases either as a single relation of arity three (triple table), as a binary relation for each property
(vertical partitioning), or as n-ary relations for entities of a given type (property tables) [Wylot et al.,
2018]. Custom (so-called native) storage techniques have also been developed for a variety of graph
models, providing efficient access for finding nodes, edges and their adjacent elements [Angles and
Gutierrez, 2008, Miller, 2013, Wylot et al., 2018]. A number of systems further allow for distributing
graphs over multiple machines based on popular NoSQL stores or custom partitioning
schemes [Wylot et al., 2018, Janke and Staab, 2018]. For further details we refer to the book chapter
by Janke and Staab [2018] and the survey by Wylot et al. [2018] dedicated to this topic.
2.2 Querying
A number of languages have been proposed for querying graphs [Angles et al., 2017], including the
SPARQL query language for RDF graphs [Harris et al., 2013]; and Cypher [Francis et al., 2018],
Table of Contents
Gremlin [Rodriguez, 2015], and G-CORE [Angles et al., 2018] for querying property graphs. We re‐
fer to Seifer et al. [2019] for an investigation of the popularity of these languages. Underlying these
query languages are some common primitives, including (basic) graph patterns, relational operators,
path expressions, and more besides [Angles et al., 2017]. We now describe these core features for
querying graphs in turn, starting with basic graph patterns.
At the core of every structured query language for graphs lie basic graph patterns [Consens and
Mendelzon, 1990, Angles et al., 2017], which follow the same model as the data graph being queried
(see Section 2.1), additionally allowing variables as terms.2 Terms in basic note
graph patterns
2 The termsare
of athus
directed edge-labelled grap
of with
divided into constants, such as Arica or venue, and variables, which we prefix a property graph
question are its ids, labels, propertie
marks,
nodes).
such as ?event or ?rel. A basic graph pattern is then evaluated against the data graph by generating
mappings from the variables of the graph pattern to constants in the data graph such that the image
of the graph pattern under the mapping (replacing variables with the assigned constants) is con‐
tained within the data graph.
Figure 2.5 provide an example of a basic graph pattern looking for the venues of Food Festivals,
along with the possible mappings generated by the graph pattern against the data graph of
Figure 2.1. In some of the presented mappings (the last two listed), multiple variables are mapped to
the same term, which may or may not be desirable depending on the application. Hence a number of
semantics have been proposed for evaluating basic graph patterns [Angles et al., 2017], amongst
which the most important are: homomorphism-based semantics, which allows multiple variables to be
mapped to the same term such that all mappings shown in Figure 2.5 would be considered results;
and isomorphism-based semantics, which requires variables on nodes and/or edges to be mapped to
unique terms, thus excluding the latter three mappings of Figure 2.5 from the results. Different lan‐
guages may adopt different semantics for evaluating basic graph patterns; for example, SPARQL
adopts a homomorphism-based semantics, while Cypher adopts an isomorphism-based semantics
specifically on edges (while allowing multiple variables to map to one node).
Figure 2.5: basic directed edge-labelled graph pattern (left ) with mappings generated over the directed
edge-labelled graph of Figure 2.1 (right)
Give us feedback!
https://kgbook.org 15/150
2/15/23, 5:13 PM Knowledge Graphs
As we will see in later examples (particularly Figure 2.7), basic graph patterns may also form cycles
(be they directed or undirected), and may replace edge labels with variables. Basic graph patterns in
the context of other models – such as property graphs – can be defined analogously by allowing vari‐
ables to replace constants in any position of the model.
We formalise basic graph patterns first for directed edge-labelled graphs, and subsequently for
property graphs [Angles et al., 2017]. For these definitions, we introduce a countably infinite set
of variables Var ranging over (but disjoint from: Con ∩ Var = ∅) the set of constants. We refer
generically to constants and variables as terms, denoted and defined as Term = Con ∪ Var. We
define a basic graph pattern for a model by simply replacing constants with terms (that may be
variables). Though we focus on directed edge-labelled graphs and property graphs, basic graph
patterns for other graph models can be defined analogously.
the set V contains the constant Food Festival and variables ?event, ?ven1 and ?ven2;
the set E contains four edges, including (?event, type, Food Festival);
the set L contains the constants type and venue.
Towards defining the results of evaluating a basic graph pattern over a data graph (following
the same model), we first define a partial mapping μ : Var → Con from variables to constants,
whose domain (the set of variables for which it is defined) is denoted by dom(μ). Given a basic
graph pattern Q, let Var(Q) denote the set of all variables appearing in (some recursively nest‐
ed element of) Q. We further denote by μ(Q) the image of Q under μ, meaning that any variable
v ∈ Var(Q) ∩ dom(μ) is replaced in Q by μ(v). Observe that when Var(Q) ⊆ dom(μ), then
μ(Q) is a data graph (in the corresponding model of Q).
Next, we define the notion of containment between data graphs. For two directed edge-labelled
graphs G1 = (V1 , E1 , L1 ) and G2 = (V2 , E2 , L2 ), we say that G1 is a sub-graph of G2 , denoted G1 ⊆
ten be defined without edges. For two property graphs G1 = (V1 , E1 , L1remark , P1 , U1 ,that
e 1 , l 1G ⊆ and
, p11 ) G2 and G2 ⊆ G1 : the former h
latter G
G2 = (V2 , E2 , L2 , P2 , U2 , e2 , l2 , p2 ), we say that G1 is a sub-graph of G2 , denoted has⊆aGnode without an incident edge. In c
1 2 , if and
cases of nodes or labels without edges cannot oc
E2 holds.
Give us feedback!
https://kgbook.org 16/150
2/15/23, 5:13 PM Knowledge Graphs
Example 2.6 — Figure 2.5 enumerates all of the mappings given by the evaluation of the depicted ba‐
sic graph pattern over the data graph of Figure 2.1. Each non-header row indicates a mapping μ.
The final results of evaluating a basic graph pattern may vary depending on the choice of se‐
mantics: the results under homomorphism-based semantics are defined as Q(G). Conversely, under
isomorphism-based semantics, mappings that send two edge variables to the same constant
and/or mappings that send two node variables to the same constant may be excluded from the
results. Henceforth we assume the more general homomorphism-based semantics.
A (basic) graph pattern transforms an input graph into a table of results (as shown in Figure 2.5). We
may then consider using the relational algebra to combine and/or transform such tables, thus form‐
ing more complex queries from one or more graph patterns. Recall that the relational algebra con‐
sists of unary operators that accept one input table, and binary operators that accept two input ta‐
Table of Contents
bles. Unary operators include projection (π ) to output a subset of columns, selection (σ ) to output a
subset of rows matching a given condition, and renaming of columns (ρ). Binary operators include
union (∪) to merge the rows of two tables into one table, difference (−) to remove the rows from the
first table present in the second table, and joins (⋈) to extend the rows of one table with rows from
the other table that satisfy a join condition. Selection and join conditions typically include equalities
(=), inequalities (≤), negation (¬), disjunction (∨), etc. From these operators, we can further define
other (syntactic) operators, such as intersection (∩) to output rows in both tables, anti-join (⊳, aka
minus) to output rows from the first table for which there are no join-compatible rows in the second
table, left-join (⟕, aka optional) to perform a join but keeping rows from the first table without a
compatible row in the second table, etc.
Basic graph patterns can then be expressed in a subset of relational algebra (namely π , σ , ρ, ⋈).
Assuming, for example, a single ternary relation G(s, p, o) representing a graph – i.e., a table G with
three columns s, p, o – the query of Figure 2.5 can be expressed in relational algebra as:
where ⋈ denotes a natural join, meaning that equality is checked across pairs of columns with the
same name in both tables (here, the join is thus performed on the subject column s). The result of
this query is a table with a column for each variable: ev, vn1, vn2. However, not all queries using
π, σ, ρ and ⋈ on G can be expressed as basic graph patterns; for example, we cannot choose which
variables to project in a basic graph pattern, but rather must project all variables not fixed to a
constant.
Graph query languages such as SPARQL [Harris et al., 2013] and Cypher [Francis et al., 2018] allow
the full use of relational operators over the results of graph patterns, giving rise to complex graph pat‐
terns [Angles et al., 2017]. Figure 2.6 presents an example of a complex graph pattern with projected
variables in bold, choosing particular variables to appear in the final results. In Figure 2.7, we give
another example of a complex graph pattern looking for food festivals or drinks festivals not held in
Santiago, optionally returning their start date and name (where available).
Figure 2.6: Complex graph pattern (left ) with mappings generated over the graph of Figure 2.1 (right)
Give us feedback!
https://kgbook.org 17/150
2/15/23, 5:13 PM Knowledge Graphs
Complex graph patterns can give rise to duplicate results; for example, the first result in Figure 2.6
appears twice since ?city1 matches Arica and ?city2 matches Viña del Mar in one result, and
vice-versa in the other. Query languages then offer two semantics: bag semantics preserves duplicates
according to the multiplicity of the underlying mappings, while set semantics (typically invoked with
a DISTINCT keyword) removes duplicates from the results.
Figure 2.7: Complex graph pattern (Q) with mappings generated (Q(G)) over the graph of Figure 2.1 (G)
If Q is a complex graph pattern, and R is a selection condition with Boolean and equality
connectives (∧, ∨, ¬, =) , then σR (Q) is a complex graph pattern.
We now define the evaluation of complex graph patterns. Given a mapping μ, for a set of vari‐
ables V ⊆ Var let μ[V] denote the mapping μ′ such that dom(μ′ ) = dom(μ) ∩ V and μ′ (v) = μ(v)
for all v ∈ dom(μ′ ) (in other words, μ[V] projects the variables V from μ). Letting R denote a
Boolean selection condition and μ a mapping, we denote by μ ⊨ R that μ satisfies the Boolean
condition. Finally, we define two mappings μ1 and μ2 to be compatible, denoted μ1 ∼ μ2 , if and
only if μ1 (v) = μ2 (v) for all v ∈ dom(μ1 ) ∩ dom(μ2 ) (i.e., they map common variables to the
Based on these operators, we can define some additional syntactic operators, such as the left-join
⟕, aka optional):
Example 2.7 — Figure 2.7 illustrates a complex graph pattern and its evaluation. Give us feedback!
https://kgbook.org 18/150
2/15/23, 5:13 PM Knowledge Graphs
2.2.3 Navigational graph patterns
A key feature that distinguishes graph query languages is the ability to include path expressions in
queries. A path expression r is a regular expression that allows for matching arbitrary-length paths
between two nodes using a regular path query (x, r, y), where x and y can be variables or constants
(or even the same term). The base path expression is where r is a constant (an edge label).
Furthermore if r is a path expression, then r∗ (Kleene star: zero-or-more) is also a path expression.
Finally, if r1 and r2 are path expressions, then r1 ∣ r2 (disjunction) and r1 ⋅ r2 (concatenation) are also
path expressions. A related notion is that of 2-way regular path queries, which also allow for querying
inverse paths; specifically, if r is path expression, then it is a 2-way path expression, and if r is a 2-
way path expression, then r− (inverse) is a 2-way path expression. Henceforth we will refer generically
to both the 1-way and 2-way variants as path expressions and regular path queries.
Regular path queries can be evaluated under a number of different semantics. For example, (Arica,
bus*, ?city) evaluated against the graph of Figure 2.1 may match the paths shown in Figure 2.8.
In fact, since a cycle is present, an infinite number of paths are potentially matched. For this reason,
Table of Contents
restricted semantics are often applied, returning only the shortest paths, or paths without repeated
nodes or edges (as in the case of Cypher).4 Rather than returning paths, anothernote option is to instead
4 Mapping variables to paths requires s
Cypher
return the (finite) set of pairs of nodes connected by a matching path (as in the case of[Francis
SPARQLet1.1).
al., 2018] returns a string
functions such as length(·) can be applied. G
hand, allows for returning paths, and supports
projecting them as graphs, applying cost functio
⋯
Figure 2.8: Example paths matching (Arica, bus*, ?city) over the graph of Figure 2.1
Regular path queries can then be used in basic graph patterns to express navigational graph pat‐
terns [Angles et al., 2017], as shown in Figure 2.9, which illustrates a query searching for food festi‐
vals in cities reachable (recursively) from Arica by bus or flight. Furthermore, when regular path
queries and graph patterns are combined with operators such as projection, selection, union, differ‐
ence, and optional, the result is known as complex navigational graph patterns [Angles et al., 2017].
then:
We now define the evaluation of a path expression on a directed-edge labelled graph under the
SPARQL 1.1-style semantics whereby the endpoints (pairs of start and end nodes) of the path are
returned [Harris et al., 2013].
r∗ [G] = V ∪ ⋃ rn [G]
n∈N+
https://kgbook.org 19/150
2/15/23, 5:13 PM Knowledge Graphs
The evaluation of a path expression on a property graph G = (V , E, L, P , U , e, l, p) can be de‐
fined analogously by adapting the first definition (in the case that r ∈ Con) as follows:
Query languages may support additional operators, some of which are syntactic (e.g., r+ is
sometimes used for one-or-more, but can be rewritten as r ⋅ r∗ ), while others may add expressivi‐
ty such as the case of SPARQL [Harris et al., 2013], which allows a limited form of negation in
expressions (e.g., !r, with r being a constant or the inverse of a constant, matching any path not
labelled r).
Var denote variables. Then the evaluation of a regular path query is defined as follows:
where μ∅ denotes the empty mapping such that dom(μ) = ∅ (the join identity).
The definition of the evaluation of a navigational graph pattern then follows from the previous
definition of a join and the definition of the evaluation of a regular path query (for a directed
edge-labelled graph or a property graph, respectively). Likewise, complex navigational graph pat‐
terns – and their evaluation – are defined by extending this definition in the natural way with
the same operators from Definition 2.8 following the same semantics seen in Definition 2.9.
Figure 2.9: Navigational graph pattern (left ) with mappings generated over the graph of Figure 2.1
(right)
Thus far, we have discussed features that form the practical and theoretical foundation of any query
language for graphs [Angles et al., 2017]. However, specific query languages for graphs may support
other features, such as aggregation (GROUP BY, COUNT, etc.), more complex filters and datatype op‐
erators (e.g., range queries on years extracted from a date), federation for querying remotely hosted
graphs over the Web, languages for updating graphs, support for entailment, etc. For more informa‐
tion, we refer to the documentation of the respective query languages (e.g., [Harris et al., 2013,
Angles et al., 2018]) and to the survey by Angles et al. [2017].
Give us feedback!
https://kgbook.org 20/150
2/15/23, 5:13 PM Knowledge Graphs
2.2.5 Query Interfaces
Knowledge graphs are often queried by non-expert users who may not be able to express their infor‐
mation needs in terms of a particular graph query language. Different types of interfaces have thus
been proposed in order to assist users in querying data graphs. Such interfaces may support, for
example:
Faceted browsing:
Users start by specifying a simple search, such as a keyword search, a type of node like Food
Festival, or possibly other kinds of search. They are then presented with a set of matching
results, and a set of facets, which are typically attributes (e.g., venue) and values (e.g., Santa
Lucía) present in the current results set. Selecting a value for a facet restricts the current re‐
sults set to include only results with the indicated value; this selection process can be applied
iteratively to restrict results per multiple facets. Often the faceted criteria are translated into
and evaluated as graph queries. Though relatively intuitive for users, such systems typically
support acyclic queries that generate lists of results (analogous to graph queries that project a
Table of Contents
single variable), and rarely support more expressive queries. Examples of faceted browsing sys‐
tems for graphs include VisiNav [Harth, 2010], Broccoli [Bast and Buchhold, 2013],
SemFacet [Arenas et al., 2016], GraFa [Moreno-Vega and Hogan, 2018], etc.
Query building:
Users are provided with a form or graphical interface that can be used to specify a graph query
without needing to understand the syntax of a specific query language. Such query builders al‐
low for incrementally adding nodes or edges to the query, assisted by features such as auto-
completion, previewing intermediate results, and graph navigation. Query builders typically al‐
low for expressing queries equivalent to (cyclic) basic graph patterns, but may not support
more expressive features of query languages as described herein. Graph query builder systems
include Smeagol [Clemmer and Davies, 2011], QueryVOWL [Haag et al., 2015], VIIQ [Jayaram
et al., 2015a], Sparklis [Ferré, 2017], RDF Explorer [Vargas et al., 2019], and more besides.
Query-by-example:
Users provide examples of positive and sometimes negative answers to their queries. For exam‐
ple, they may provide as positive examples the nodes Arica , Santiago , Viña del Mar , and as neg‐
ative examples the nodes Chile , Lima , where the system will then “reverse engineer” a query
that returns positive examples but not negative examples (in this case, the query proposed may
return nodes of type City whose country is Chile). Query-by-example systems typically
support basic graph patterns, and may not support more expressive querying features. They are
useful in cases where users have examples of what they are looking for, but are not necessarily
sure of the query they need to retrieve similar examples. Query-by-example systems for graphs
include GQBE [Jayaram et al., 2015b] and SPARQLByE [Diaz et al., 2016].
Question answering:
Users express their queries as questions in natural language; for example, they might ask “What
food festivals will be held in Arica?”. The question answering system will then generate answers
from the graph based on its best interpretation of the question. We identify three types of ques‐
tion answering system. Navigation-based systems identify entities/nodes from the graph that are
mentioned in the query, and then attempt to navigate edges from those nodes whose labels best
match the question; for example, they may match the nodes Food Festival and Arica in the
graph based on the question, and from there, try to navigate edges in the graph whose labels
match the question in order to find answers. Template-based systems rather pre-suppose a fixed
list of question templates expressed in the query language, with placeholder variables that will
be replaced with entities/nodes detected in the question; a template matched for the previous
example may be of the form “What X will be held in Y?”. Translation-based systems attempt to
translate the question into a query in the structured query language, using (typically neural)
machine translation techniques. The latter two types of question answering systems can addi‐
tionally return a graph query that explains the answers generated. Question answering systems
are often very intuitive to use, but may not always return correct results, particularly when
considering complex questions/queries. Examples of question answering systems for knowledge
graphs include Treo [Freitas et al., 2011], NFF [Hu et al., 2018], TemplateQA [Zheng et al.,
2018], WDAqua-core1 [Diefenbach et al., 2020], and more besides.
Give us feedback!
https://kgbook.org 21/150
2/15/23, 5:13 PM Knowledge Graphs
Such query interfaces enable non-expert users to formulate queries over graphs, which in turn broad‐
ens the potential impact of knowledge graphs.
Chapter 3
Chapter 2. We refer to a knowledge graph as a data graph potentially enhanced with representations
of schema, identity, context, ontologies and/or rules. These additional representations may be em‐
bedded in the data graph, or layered above. Representations for schema, identity and context are dis‐
cussed now, while ontologies and rules will be discussed in Chapter 4.
3.1 Schema
One of the benefits of modelling data as graphs – versus, for example, the relational model – is the
option to forgo or postpone the definition of a schema. However, when modelling data as graphs,
schemata can be used to prescribe a high-level structure and/or semantics that the graph follows or
should follow. We discuss three types of graph schemata: semantic, validating, and emergent.
A semantic schema allows for defining the meaning of high-level terms (aka vocabulary or terminolo‐
gy) used in the graph, which facilitates reasoning over graphs using those terms. Looking at
Figure 2.1, for example, we may notice some natural groupings of nodes based on the types of enti‐
ties to which they refer. We may thus decide to define classes, such as Event, City, etc., to denote
these groupings. In fact, Figure 2.1 already illustrates three low-level classes – Open Market, Food
Market, Drinks Festival – grouping similar entities with an edge labelled type. We may subse‐
quently wish to capture some relations between some of these classes. In Figure 3.1, we present a
class hierarchy for events where children are defined to be sub-classes of their parents such that if we
find an edge EID15 –type ➛ Food Festival in our graph, we may also infer that EID15 –type ➛ Festival and
EID15 –type ➛ Event hold in the graph.
Aside from classes, we may also wish to define the semantics of edge labels, aka properties. Returning
to Figure 2.1, we may consider that the properties city and venue are sub-properties of a more general
property location, such that given an edge Santa Lucía –city ➛ Santiago , for example, we may also infer
that Santa Lucía –location ➛ Santiago must hold as an edge in the graph. We may also consider, for ex‐
ample, that bus and flight are both sub-properties of a more general property connects to. Along these
Give
lines, properties may also form a hierarchy similar to what we saw for classes. We may further define us feedback!
https://kgbook.org 22/150
2/15/23, 5:13 PM Knowledge Graphs
the domain of properties, indicating the class(es) of entities for nodes from which edges with that
property extend; for example, we may define that the domain of connects to is a class Place, such
that given the previous sub-property relations, we infer Arica –type ➛ Place . Conversely, we may de‐
fine the range of properties, indicating the class(es) of entities for nodes to which edges with that
property extend; for example, we may define that the range of city is a class City, inferring that
Arica –type ➛ City .
A prominent standard for defining a semantic schema for (RDF) graphs is the RDF Schema (RDFS)
standard [Brickley and Guha, 2014], which allows for defining sub-classes, sub-properties, domains,
and ranges amongst the classes and properties used in an RDF graph, where such definitions can be
serialised as a graph. We illustrate the semantics of these features in Table 3.1 and provide a con‐
crete example of definitions in Figure 3.2 for a sample of terms used in the running example. These
definitions can then be embedded into a data graph. More generally, the semantics of terms used in a
graph can be defined in much more depth than seen here, as is supported by the Web Ontology
Language (OWL) standard [Hitzler et al., 2012] for RDF graphs. We will return to such semantics lat‐
er in Chapter 4.
Table of Contents
Figure 3.2: Example schema with sub-classes, sub-properties, domains, and ranges
Semantic schemata are typically defined for incomplete graph data, where the absence of an edge
between two nodes, such as Viña del Mar –flight ➛ Arica , does not mean that the relation does not hold
in the real world. Therefore, from the graph of Figure 2.1, we cannot assume that there is no flight
between Viña del Mar and Arica. In contrast, if the Closed World Assumption (CWA) were adopted –
as is the case in many classical database systems – it would be assumed that the data graph is a com‐
plete description of the world, thus allowing to assert with certainty that no flight exists between the
two cities. Systems that do not adopt the CWA are said to adopt the Open World Assumption (OWA).
Considering our running example, it would be unreasonable to assume that the tourism organisation
has complete knowledge of everything describable in its knowledge graph, and hence adopting the
OWA appears more appropriate. However, it can be inconvenient if a system is unable to definitely
answer “yes” or “no” to questions such as “is there a flight between Arica and Viña del Mar?”, especially
when the organisation is certain that it has complete knowledge of the flights. A compromise be‐
tween OWA and CWA is the Local Closed World Assumption (LCWA), where portions of the data graph
are assumed to be complete.
When graphs are used to represent diverse, incomplete data at large scale, the OWA is the most ap‐
propriate choice for a default semantics. But in some scenarios, we may wish to guarantee that our
data graph – or specific parts thereof – are in some sense “complete”. Returning to Figure 2.1, for ex‐
ample, we may wish to ensure that all events have at least a name, a venue, a start date, and an end
date, such that applications using the data – e.g., one that sends event notifications to users – can en‐
Give us feedback!
sure that they have the minimal information required. Furthermore, we may wish to ensure that the
https://kgbook.org 23/150
2/15/23, 5:13 PM Knowledge Graphs
city of an event is stated to be a city (rather than inferring that it is a city). We can define such con‐
straints in a validating schema and validate the data graph with respect to the resulting schema, list‐
ing constraint violations (if any). Thus while semantic schemata allow for inferring new graph data,
validating schemata allow for validating a given data graph with respect to some constraints.
A standard way to define a validating schema for graphs is using shapes [Knublauch and
Kontokostas, 2017, Prud'hommeaux et al., 2014, Labra Gayo et al., 2018]. A shape targets a set of
nodes in a data graph and specifies constraints on those nodes. The shape’s target can be defined in
many ways, such as targeting all instances of a class, the domain or range of a property, the result of
a query, nodes connected to the target of another shape by a given property, etc. Constraints can
then be defined on the targeted nodes, such as to restrict the number or types of values taken on a
given property, the shapes that such values must satisfy, etc.
A shapes graph is formed from a set of interrelated shapes. Shapes graphs can be depicted as UML-
like class diagrams, where Figure 3.3 illustrates an example of a shapes graph based on Figure 2.1,
defining constraints on four interrelated shapes. Each shape – denoted with a box like Place , Event ,
etc. – is associated with a set of constraints. Nodes conform to a shape if and only if they satisfy all
Table of Contents
constraints defined on the shape. Inside each shape box are placed constraints on the number (e.g.,
[1..*] denotes one-to-many, [1..1] denotes precisely one, etc.) and types (e.g., string, date‐
Time, etc.) of nodes that conforming nodes can relate to with a property (e.g., name, start, etc.).
Another option is to place constraints on the number of nodes conforming to a particular shape that
the conforming node can relate to with a property (thus generating edges between shapes); for exam‐
ple, Event – 1..* ➛Venue denotes that conforming nodes for Event must relate to at least one node
venue
with the property venue that conforms to the Venue shape. Shapes can inherit the constraints of par‐
ent shapes – with inheritance denoted with an △ connector – as in the case of City and Venue ,
whose conforming nodes must also conform to the Place shape.
Given a shape and a targeted node, it is possible to check if the node conforms to that shape or not,
which may require checking conformance of other nodes; for example, the node EID15 conforms to
the Event shape not only based on its local properties, but also based on conformance of Santa Lucía
to Venue and Santiago to City . Conformance dependencies may also be recursive, where the confor‐
mance of Santiago to City requires that it conforms to Place , which requires that Viña del Mar and
Arica conform to Place , and so on. Conversely, EID16 does not conform to Event , as it does not
have the start and end properties required by the example shapes graph.
When declaring shapes, the data modeller may not know in advance the entire set of properties that
some nodes can have (now or in the future). An open shape allows the node to have additional prop‐
erties not specified by the shape, while a closed shape does not. For example, if we add the edge
Santiago –founder ➛ Pedro de Valdivia to the graph represented in Figure 2.1, then Santiago only conforms
to the City shape if the shape is defined as open (since the shape does not mention founder).
Practical languages for shapes often support additional Boolean features, such as conjunction (and),
disjunction (or), and negation (not) of shapes; for example, we may say that all the values of venue
should conform to the shape Venue and (not City) , making explicit that venues in the data graph
should not be directly given as cities. However, shapes languages that freely combine recursion and
Give us feedback!
https://kgbook.org 24/150
2/15/23, 5:13 PM Knowledge Graphs
negation may lead to semantic problems, depending on how their semantics are defined. To illus‐
trate, consider the following case inspired by the barber paradox [Labra Gayo et al., 2018], involving
a shape Barber whose conforming nodes shave at least one node conforming to
Person and (not Barber) . Now, given (only) Bob –shave ➛ Bob with Bob conforming to Person , does
Bob conform to Barber ? If yes – if Bob conforms to Barber – then Bob violates the constraint by
not shaving at least one node conforming to Person and (not Barber) . If no – if Bob does not conform
to Barber – then Bob satisfies the Barber constraint by shaving such a node. Semantics to avoid
such paradoxical situations have been proposed based on stratification [Boneva et al., 2017], partial
assignments [Corman et al., 2018], and stable models [Gelfond and Lifschitz, 1988].
Although validating schemata and semantic schemata serve different purposes, they can complement
each other. In particular, a validating schema can take into consideration a semantic schema, such
that, for example, validation is applied on the data graph including inferences. Taking the class hier‐
archy of Figure 3.1 and the shapes graph of Figure 3.3, for example, we may define the target of the
Event shape as the nodes that are of type Event (the class). If we first apply inferencing with re‐
spect to the class hierarchy of the semantic schema, the Event shape would now target EID15 and
Table of Contents
EID16 . The presence of a semantic schema may, however, require adapting the validating schema.
Taking into account, for example, the aforementioned class hierarchy would require defining a re‐
laxed cardinality on the type property. Open shapes may also be preferred in such cases rather than
enumerating constraints on all possible properties that may be inferred on a node.
Two shapes languages have recently emerged for RDF graphs: Shape Expressions (ShEx), published as
a W3C Community Group Report [Prud'hommeaux et al., 2014]; and SHACL (Shapes Constraint
Language), published as a W3C Recommendation [Knublauch and Kontokostas, 2017]. These lan‐
guages support the discussed features (and more) and have been adopted for validating graphs in a
number of domains relating to healthcare [Thornton et al., 2019], scientific literature [Hammond et
al., 2017], spatial data [Car et al., 2019], amongst others. More details about ShEx and SHACL can
be found in the book by Labra Gayo et al. [2018]. A recently proposed language that can be used as
a common basis for both ShEx and SHACL reveals their similarities and differences [Labra Gayo et
al., 2019]. A similar notion of schema has been proposed by Angles [2018] for property graphs.
We formally define shapes following the conventions of Labra Gayo et al. [2019].
∣ ¬ϕ negation of shape ϕ
∣ @s reference to shape with label s
ϕ{min, max} between min and max outward edges (inclusive) with label p
p
∣
Example 3.1 — The shapes schema from Figure 3.3 can be expressed as:
type venue
∧ ⊤{1, ∗}∧ Venue {1, ∗}
indoor city
Venue ↦ Place ∧ Δboolean {0, 1}∧
City {0, 1}
population
City ↦ Place ∧ (Δint ∧ Ψ>5000 ){0, 1}
Give us feedback!
https://kgbook.org 25/150
2/15/23, 5:13 PM Knowledge Graphs
lat long
Place ↦
Δfloat {0, 1}∧
Δfloat {0, 1}
f light bus
∧ Place {0, ∗}∧ Place {0, ∗}
For example, Event is a shape label (an element of S ) that maps to a shape (an element of ϕ). This
mapping is defined by λ.
In a shapes schema, shapes may refer to other shapes, giving rise to a graph that is sometimes
known as the shapes graph [Knublauch and Kontokostas, 2017]. Figure 3.3 illustrates a shapes
graph of this form.
The semantics of a shape is defined in terms of the evaluation of that shape over each node of a
given data graph. The semantics of a shapes schema, in turn, is the result of evaluating each
shape of the schema over each node of a given data graph; the result of this evaluation is a
shapes map.
The shapes map σ is a way of labelling the nodes of G with the labels of shapes from S . If
σ(v, s) = 1, then node v is labelled s (possibly amongst other labels); otherwise if σ(v, s) = 0,
then node v is not labelled s. The precise semantics depends on whether or not σ is a total or
partial mapping: whether or not it is defined for every pair in V × S . Herein we present the se‐
mantics for the more straightforward case wherein σ is assumed to be a total shapes map.
[¬ϕ] G,v,σ
= 1 − [ϕ]G,v,σ
[@s]G,v,σ = 1 iff σ(v, s) = 1
ϕ{min, max}]G,v,σ = 1 iff min ≤ ∣{(v, p, u) ∈ E ∣ [ϕ]G,u,σ = 1}∣ ≤ max
p
[
Typically for the purposes of validating a graph with respect to a shapes schema, a target is de‐
fined that requires certain nodes to satisfy certain shapes.
The nodes that a shape targets can be selected a manual selection, based on the type(s) of the
nodes, based on the results of a graph query, etc. [Corman et al., 2018, Labra Gayo et al., 2019].
Lastly, we define the notion of a valid graph under a given shapes schema and target based on
the existence of a shapes map satisfying certain conditions.
Give us feedback!
https://kgbook.org 26/150
2/15/23, 5:13 PM Knowledge Graphs
Example 3.2 — Taking the graph G from Figure 2.1 and the shapes schema Σ from Figure 3.3, first
assume an empty shapes target T = {}. If we consider a shapes map where (e.g.) σ( EID15 , Event ) =
1, σ( Santa Lucía , Venue ) = 1, σ( Santa Lucía , Place ) = 1, etc., but where σ( EID16 , Event ) = 0 (as it
does not have the required values for start and end), etc., then we see that G is valid under Σ and T .
However, if we were to define a shapes target T to ensure that the Event shape targets EID15 and
EID16 – i.e., to define T such that {( EID15 , Event ), ( EID16 , Event )} ⊆ T – then the graph would
no longer be valid under Σ and T since EID16 does not satisfy Event .
The semantics we present here assumes that each node in the graph either satisfies or does not
satisfy each shape labelled by the schema. More complex semantics – for example, based on
Kleene’s three-valued logic [Corman et al., 2018, Labra Gayo et al., 2019] – have been proposed
that support partial shapes maps, where the satisfaction of some nodes for some shapes can be
left as undefined. Shapes languages in practice may support other more advanced forms of con‐
straints, such as counting on paths [Knublauch and Kontokostas, 2017]. In terms of implement‐
ing validation with respect to shapes, work has been done on translating constraints into sets of
graph queries, whose results are input to a SAT solver for recursive cases [Corman et al., 2019].
Table of Contents
Both semantic and validating schemata require a domain expert to explicitly specify definitions and
constraints. However, a data graph will often exhibit latent structures that can be automatically ex‐
tracted as an emergent schema [Pham et al., 2015] (aka graph summary [Liu et al., 2018, Čebirić et al.,
2019, Spahiu et al., 2016]).
A framework often used for defining emergent schema is that of quotient graphs, which partition
groups of nodes in the data graph according to some equivalence relation while preserving some
structural properties of the graph. Taking Figure 2.1, we can intuitively distinguish different types of
nodes based on their context, such as event nodes, which link to venue nodes, which in turn link to
city nodes, and so forth. In order to describe the structure of the graph, we could consider six parti‐
tions of nodes: event, name, venue, class, date-time, city. In practice, these partitions may be computed
based on the class or shape of the node. Merging the nodes of each partition into one node while pre‐
serving edges leads to the quotient graph shown in Figure 3.4: the nodes of this quotient graph are
the partitions of nodes from the data graph and an edge X –y ➛ Z is included the quotient graph if
and only if there exists x ∈ X and z ∈ Z such that x –y ➛ z is in the original data graph.
Figure 3.4: Example quotient graph simulating the data graph in Figure 2.1
There are many ways in which quotient graphs may be defined, depending not only on how nodes
are partitioned, but also how the edges are defined. Different quotient graphs may provide different
guarantees with respect to the structure they preserve. Formally, we can say that every quotient
graph simulates its input graph (based on the simulation relation of set membership between data
nodes and quotient nodes), meaning that for all x ∈ X with x an input node and X a quotient node,
if x –y ➛ z is an edge in the data graph, then there must exist an edge X –y ➛ Z in the quotient
graph such that z ∈ Z ; for example, the quotient graph of Figure 3.4 simulates the data graph of
Figure 2.1. However, this quotient graph seems to suggest (for instance) that EID16 would have a
start and end date in the data graph when this is not the case. A stronger notion of structural preser‐
vation is given by bisimilarity, which in this case would further require that if X –y ➛ Z is an edge
in the quotient graph, then for all x ∈ X , there must exist a z ∈ Z such that x –y ➛ z is in the data
Give us feedback!
https://kgbook.org 27/150
2/15/23, 5:13 PM Knowledge Graphs
graph; this is not satisfied by EID16 in the quotient graph of Figure 3.4, which does not have an out‐
going edge labelled start or end in the original data graph. Figure 3.5 illustrates a bisimilar version of
the quotient graph, splitting the event partition into two nodes reflecting their different outgoing
edges. An interesting property of bisimilarity is that it preserves forward-directed paths: given a path
expression r without inverses and two bisimilar graphs, r will match a path in one graph if and only
if it matches a corresponding path in the other bisimilar graph. One can verify, for example, that a
path matches x –city⋅(flight|bus)* ➛ z in Figure 2.1 if and only if there is a path matching X –
city⋅(flight|bus)* ➛ Z in Figure 3.5 such that x ∈ X and z ∈ Z .
Table of Contents
Figure 3.5: Example quotient graph bisimilar with the data graph in Figure 2.1
There are many ways in which quotient graphs may be defined, depending on the equivalence rela‐
tion that partitions nodes. Furthermore, there are many ways in which other similar or bisimilar
graphs can be defined, depending on the (bi)simulation relation that preserves the data graph’s
structure [Čebirić et al., 2019]. Such techniques aim to summarise the data graph into a higher-level
topology. In order to reduce the memory overhead of the quotient graph, in practice, nodes may
rather be labelled with the cardinality of the partition and/or a high-level label (e.g., event, city) for
the partition rather than storing the labels of all nodes in the partition.
Various other forms of emergent schema not directly based on a quotient graph framework have also
been proposed; examples include emergent schemata based on relational tables [Pham et al., 2015],
and baseed on formal concept analysis [González and Hogan, 2018]. Emergent schemata may be
used to provide a human-understandable overview of the data graph, to aid with the definition of a
semantic or validating schema, to optimise the indexing and querying of the graph, to guide the inte‐
gration of data graphs, and so forth. We refer to the survey by Čebirić et al. [2019] dedicated to the
topic for further details.
V is a partition of V without the empty set, i.e., V ⊆ (2V − ∅), V = ⋃U ∈V U , and for all
A quotient graph can “merge” multiple nodes into one node, keeping the edges of its constituent
nodes. For an input graph G = (V , E, L), there is an exponential number of possible quotient
graphs based on partitions of the input nodes. On one extreme, the input graph is a quotient
graph of itself (turning nodes like u into singleton nodes like {u} ). On the other extreme, a
single node V , with all input nodes, and loops (V , l, V ) for each edge-label l used in the set of
input edges E , is also a quotient graph. Quotient graphs typically fall somewhere in between,
where the partition V of V is often defined in terms of an equivalence relation ∼ on the set V
such that V := ∼/V ; i.e., V is defined as the quotient set of V with respect to ∼; for example, we
might define an equivalence relation on nodes such that u ∼ v if and only if they have the same
set of defined types, where ∼/V is then a partition whose parts contain all nodes with the same Give us feedback!
https://kgbook.org 28/150
2/15/23, 5:13 PM Knowledge Graphs
types. Another way to induce a quotient graph is to define the partition in a way that preserves
some of the topology (i.e., connectivity) of the input graph. One way to formally define this idea
is through simulation and bisimulation.
Bisimulation (≈) is then an equivalence relation on graphs. By defining the (bi)simulation rela‐
tion R in terms of set membership ∈, every quotient graph simulates its input graph, but does
not necessarily bisimulate its input graph. This gives rise to the notion of bisimilar quotient
graphs.
Example 3.3 — Figures 3.4 and 3.5 exemplify quotient graphs for the graph of Figure 2.1. Figure 3.4
simulates but is not bisimilar to the data graph. Figure 3.5 is bisimilar to the data graph. Often the goal
will be to compute the most concise quotient graph that satisfies a given condition; for example, the
nodes without outgoing edges in Figure 3.5 could be merged while preserving bisimilarity.
3.2 Identity
Figure 2.1 uses nodes like Santiago , but to which Santiago does this node refer? Do we refer to
Santiago de Chile, Santiago de Cuba, Santiago de Compostela, or do we perhaps refer to the indie
rock band Santiago? Based on edges such as Santa Lucía –city ➛ Santiago , we may deduce that it is one
of the three cities mentioned (not the rock band), and based on the fact that the graph describes
tourist attractions in Chile, we may further deduce that it refers to Santiago de Chile. Without fur‐
ther details, however, disambiguating nodes of this form may rely on heuristics prone to error in more
difficult cases. To help avoid such ambiguity, first we may use globally-unique identifiers to avoid
naming clashes when the knowledge graph is extended with external data, and second we may add
external identity links to disambiguate a node with respect to an external source.
Assume we wished to compare tourism in Chile and Cuba, and we have acquired an appropriate
knowledge graph for Cuba similar to the one we have for Chile. We can merge two graphs by taking
their union. However, as shown in Figure 3.6, using an ambiguous node like Santiago may yield a
naming clash: the node is referring to two different real-world cities in both graphs, where the
merged graph indicates that Santiago is a city in both Chile and Cuba (rather than two distinct
cities).5 To avoid such clashes, long-lasting persistent identifiers (PIDs) [Hakala, 2010]
note can
5 Such be created
a naming clash is not unique to grap
in order to uniquely identify an entity; examples of PID schemes include trees, Digitaletc.
Object Identifiers
(DOIs) for papers, ORCID iDs for authors, International Standard Book Numbers (ISBNs) for books,
Alpha-2 codes for counties, and more besides.
Give us feedback!
https://kgbook.org 29/150
2/15/23, 5:13 PM Knowledge Graphs
Figure 3.6: Result of merging two graphs with ambiguous local identifiers
In the context of the Semantic Web, the RDF data model goes one step further and recommends that
global Web identifiers be used for nodes and edge labels. However, rather than adopt the Uniform
Resource Locators (URLs) used to identify the location of information resources such as webpages,
RDF 1.1 proposes to use Internationalised Resource Identifiers (IRIs) to identify non-information re‐
Table of Contents
sources such as cities or events.6 Hence, for example, in the RDF note representation of the Identifiers (URIs) can be
6 Uniform Resource
Wikidata [Vrandečić and Krötzsch, 2014] – a knowledge graph proposed to to locate information
complement Wikipedia,resources, and Uniform R
sources. Internationalised Resource Identifiers
discussed in more detail in Chapter 10 – while the URL https://www.wikidata.org/wiki/Q2887 refers
http://example.com/Ñamto a ).
webpage that can be loaded in a browser providing human-readable metadata about Santiago, the
IRI http://www.wikidata.org/entity/Q2887 refers to the city itself. Distinguishing the identifiers for the web‐
page and the city itself avoids naming clashes; for example, if we use the URL to identify both the
webpage and the city, we may end up with an edge in our graph, such as (with readable labels be‐
low the edge):
Such an edge leaves ambiguity: was Pedro de Valdivia the founder of the webpage, or the city?
Using IRIs for entities distinct from the URLs for the webpages that describe them avoids such am‐
biguous cases, where Wikidata thus rather defines the previous edge using less ambiguous identi‐
fiers, as follows:
using IRIs for the city, person, and founder of, distinct from the webpages describing them. These
Wikidata identifiers use the prefix http://www.wikidata.org/entity/ for entities and the prefix
http://www.wikidata.org/prop/direct/ for relations. Such prefixes are known as namespaces, and are of‐
ten abbreviated with prefix strings, such as wd: or wdt:, where the latter edge can then be written
more concisely using such abbreviations as th edge wd:Q2887 –wdt:P112 ➛ wd:Q203534 .
If HTTP IRIs are used to identify the graph’s entities, when the IRI is looked up (via HTTP), the web-
server can return (or redirect to) a description of that entity in formats such as RDF. This further en‐
ables RDF graphs to link to related entities described in external RDF graphs over the Web, giving
rise to Linked Data [Berners-Lee, 2006, Heath and Bizer, 2011] (discussed in Chapter 9). Though
HTTP IRIs offer a flexible and powerful mechanism for issuing global identifiers on the Web, they are
not necessarily persistent: websites may go offline, the resources described at a given location may
change, etc. In order to enhance the persistence of such identifiers, Persistent URL (PURL) services of‐
fer redirects from a central server to a particular location, where the PURL can be redirected to a
new location if necessary, changing the address of a document without changing its identifier. The
persistence of HTTP IRIs can then be improved by using namespaces defined through PURL services.
Assume that the tourist board opts to define the chile: namespace with an IRI such as
http://turismo.cl/entity/ on a web-server that they control, allowing nodes such as
chile:Santiago – a shortcut for the IRI http://turismo.cl/entity/Santiago – to be looked up over the Web.
Give us feedback!
https://kgbook.org 30/150
2/15/23, 5:13 PM Knowledge Graphs
While using such a naming scheme helps to avoid naming clashes, the use of IRIs does not necessari‐
ly help ground the identity of a resource. For example, an external geographic knowledge graph may
assign the same city the IRI geo:SantiagoDeChile in their own namespace, where we have no direct way
of knowing that the two identifiers refer to the same city. If we merge the two knowledge graphs, we
will end up with two distinct nodes for the same city, and thus not integrate their data.
There are a number of ways to ground the identity of an entity. The first is to associate the entity
with uniquely-identifying information in the graph, such as its geo-coordinates, its postal code, the
year it was founded, etc. Each additional piece of information removes ambiguity regarding which
city is being referred to, providing (for example) more options for matching the city with its ana‐
logue in external sources. A second option is to use identity links to state that a local entity has the
same identity as another coreferent entity found in an external source; an instantiation of this concept
can be found in the OWL standard, which defines the owl:sameAs property relating coreferent enti‐
ties. Using this property, we could state the edge chile:Santiago –owl:sameAs ➛ geo:SantiagoDeChile in our
RDF graph, thus establishing an identity link between the corresponding nodes in both graphs.
Rather than specifying pairwise identity links between all knowledge graphs, it suffices if two knowl‐
Table of Contents
edge graphs provide corresponding identity links to the same external knowledge graph, such as
DBpedia or Wikidata; for example, if the local knowledge graph provides an identity link to Wikidata
indicating chile:Santiago –owl:sameAs ➛ wd:Q2887 , while the remote knowledge graph has the identity
link geo:SantiagoDeChile –owl:sameAs ➛ wd:Q2887 , then we can infer chile:Santiago –owl:sameAs ➛
geo:SantiagoDeChile . The semantics of owl:sameAs defined by the OWL standard then allows us to
combine the data for both nodes. Such semantics will be discussed later in Chapter 4. Ways in which
identity links can be computed will also be discussed later in Chapter 8.
3.2.3 Datatypes
Consider the two date-times on the left of Figure 2.1: how should we assign these nodes
persistent/global identifiers? Intuitively it would not make much sense, for example, to assign IRIs to
these nodes since their syntactic form tells us what they refer to: specific dates and times in March
2020. This syntactic form is further recognisable by machine, meaning that with appropriate soft‐
ware, we could order such values in ascending or descending order, extract the year, etc.
Most practical data models for graphs allow for defining nodes that are datatype values. RDF utilises
XML Schema Datatypes (XSD) [Peterson et al., 2012], amongst others, where a datatype node is given
as a pair (l, d) where l is a lexical string, such as "2020-03-29T20:00:00", and d is an IRI denoting
the datatype, such as xsd:dateTime. The node is then denoted
"2020-03-29T20:00:00"^^xsd:dateTime . Datatype nodes in RDF are called literals and are not allowed to
have outgoing edges. Other datatypes commonly used in RDF data include xsd:string, xsd:in‐
teger, xsd:decimal, xsd:boolean, etc. If the datatype is omitted, the value is assumed to be of
type xsd:string. Applications built on top of RDF can then recognise these datatypes, parse them
into datatype objects, and apply equality checks, normalisation, ordering, transformations, etc., ac‐
cording to their standard definition. In the context of property graphs, Neo4j [Miller, 2013] also de‐
fines a set of internal datatypes on property values that includes numbers, strings, Booleans, spatial
points, and temporal values.
3.2.4 Lexicalisation
Global identifiers for entities will sometimes have a human-interpretable form, such as chile:Santiago ,
but the identifier strings themselves do not carry any formal semantic significance. In other cases,
the identifiers used may not be human-interpretable by design. In Wikidata, for instance, Santiago de
Chile is identified as wd:Q2887 , where such a scheme has the advantage of providing better persis‐
tence and of not being biased to a particular human language. As a real-world example, the Wikidata
identifier for Eswatini ( wd:Q1050 ) was not affected when the country changed its name from
Swaziland, and does not necessitate choosing between languages for creating (more readable) IRIs
such as wd:Eswatini (English), wd:eSwatini (Swazi), wd:Esuatini (Spanish), etc.
Give us feedback!
https://kgbook.org 31/150
2/15/23, 5:13 PM Knowledge Graphs
Since identifiers can be arbitrary, it is common to add edges that provide a human-interpretable la‐
bel for nodes, such as wd:Q2887 –rdfs:label ➛ “Santiago” , indicating how people may refer to the subject
node linguistically. Linguistic information of this form plays an important role in grounding knowl‐
edge such that users can more clearly identify which real-world entity a particular node in a knowl‐
edge graph actually references [de Melo, 2015]; it further permits cross-referencing entity labels
with text corpora to find, for example, documents that potentially speak of a given entity [Martínez-
Rodríguez et al., 2020]. Labels can be complemented with aliases (e.g., wd:Q2887 –skos:altLabel ➛
“Santiago de Chile” ) or comments (e.g. wd:Q2887 –rdfs:comment ➛ “Santiago is the capital of Chile” ) to further
help ground the node’s identity.
Nodes such as “Santiago” denote string literals, rather than an identifier. Depending on the specific
graph model, such literal nodes may also be defined as a pair (s, l), where s denotes the string and l
a language code; in RDF, for example we may state chile:City –rdfs:label ➛ “City”@en , chile:City –rdfs:label ➛
“Ciudad”@es , etc., indicating labels for the node in different languages. In other models, the pertinent
language can rather be specified, e.g., via metadata on the edge. Knowledge graphs with human-in‐
terpretable labels, aliases, comments, etc., (in various languages) are sometimes called (multilingual)
Table of Contents
When modelling incomplete information, we may in some cases know that there must exist a partic‐
ular node in the graph with particular relationships to other nodes, but without being able to identi‐
fy the node in question. For example, we may have two co-located events chile:EID42 and chile:EID43
whose venue has yet to be announced. One option is to simply omit the venue edges, in which case
we lose the information that these events have a venue and that both events have the same venue.
Another option might be to create a fresh IRI representing the venue, but semantically this becomes
indistinguishable from there being a known venue. Hence some graph models permit the use of exis‐
tential nodes, represented here as a blank circle:
These edges denote that there exists a common venue for chile:EID42 and chile:EID42 without identify‐
ing it. Existential nodes are supported in RDF as blank nodes [Cyganiak et al., 2014], which are also
commonly used to support modelling complex elements in graphs, such as RDF lists [Cyganiak et al.,
2014, Hogan et al., 2014]. Figure 3.7 exemplifies an RDF list, which uses blank nodes in a linked-list
structure to encode order. Though existential nodes can be convenient, their presence can complicate
operations on graphs, such as deciding if two data graphs have the same structure modulo existential
nodes [Cyganiak et al., 2014, Hogan, 2017]. Hence methods for skolemising existential nodes in
graphs – replacing them with canonical labels – have been proposed [Longley and Sporny, 2019,
Hogan, 2017]. Other authors rather call to minimise the use of such nodes in graph data [Heath and
Bizer, 2011].
Figure 3.7: RDF list representing the three largest peaks of Chile, in order
3.3 Context
Many (arguably all) facts presented in the data graph of Figure 2.1 can be considered true with re‐
spect to a certain context. With respect to temporal context, Santiago has existed as a city since 1541,
flights from Arica to Santiago began in 1956, etc. With respect to geographic context, the graph de‐
Give
scribes events in Chile. With respect to provenance, data relating to EID15 were taken from – and are us feedback!
https://kgbook.org 32/150
2/15/23, 5:13 PM Knowledge Graphs
thus said to be true with respect to – the Ñam webpage on January 4th, 2020. Other forms of context
may also be used. We may further combine contexts, such as to indicate that Arica is a Chilean city
(geographic) since 1883 (temporal) per the Treaty of Ancón (provenance).
By context we herein refer to the scope of truth, i.e., the context in which some data are held to be
true [McCarthy, 1993, Guha et al., 2004]. The graph of Figure 2.1 leaves much of its context implic‐
it. However, making context explicit can allow for interpreting the data from different perspectives,
such as to understand what held true in 2016, what holds true excluding webpages later found to
have spurious data, etc. As seen previously, context for graph data may be considered at different
levels: on individual nodes, individual edges, or sets of edges (sub-graphs). We now discuss various
representations by which context can be made explicit at different levels.
The first way to represent context is to consider it as data no different from other data. For example,
the dates for the event EID15 in Figure 2.1 can be seen as representing a form of temporal context,
Table of Contents
indicating the temporal scope within which edges such as EID15 –venue ➛ Santa Lucía are held true.
Another option is to change a relation represented as an edge, such as Santiago –flight ➛ Arica , into a
node, such as seen in Figure 2.3a, allowing us to assign additional context to the relation. While in
these examples context is represented in an ad hoc manner, a number of specifications have been
proposed to represent context as data in a more standard way. One example is the Time
Ontology [Cox et al., 2017], which specifies how temporal entities, intervals, time instants, etc. – and
relations between them such as before, overlaps, etc. – can be described in RDF graphs in an interop‐
erable manner. Another example is the PROV Data Model [Gil et al., 2013], which specifies how
provenance can be described in RDF graphs, where entities (e.g., graphs, nodes, physical document)
are derived from other entities, are generated and/or used by activities (e.g., extraction, authorship),
and are attributed to agents (e.g., people, software, organisations).
3.3.2 Reification
Often we may wish to directly define the context of edges themselves; for example, we may wish to
state that the edge Santiago –flight ➛ Arica is valid from 1956. While we could use the pattern of turn‐
ing the edge into a node – as illustrated in Figure 2.3a – to directly represent such context, another
option is to use reification, which allows for making statements about statements in a generic manner
(or in the case of a graph, for defining edges about edges). In Figure 3.8 we present three forms of
reification that can be used for modelling temporal context on the aforementioned edge within a di‐
rected edge-labelled graph [Hernández et al., 2015]. We use e to denote an arbitrary identifier repre‐
senting the edge itself to which the context can be associated. Unlike in a direct representation, e
represents an edge, not a flight. RDF reification [Cyganiak et al., 2014] (Figure 3.8a) defines a new
node e to represent the edge and connects it to the source node (via subject), target node (via
object), and edge label (via predicate) of the edge. In contrast, n-ary relations [Cyganiak et al., 2014]
(Figure 3.8b) connect the source node of the edge directly to the edge node e with the label of the
edge; the target node of the edge is then connected to e (via value). Finally, singleton
properties [Nguyen et al., 2014] (Figure 3.8c) rather use e as an edge label, connecting it to a node
indicating the original edge label (via singleton). Other forms of reification have been proposed in the
literature, including, for example, NdFluents [Giménez-García et al., 2017]. In general, a reified edge
does not assert the edge it reifies; for example, we may reify an edge to state that it is no longer
valid. We refer to Hernández et al. [2015] for further comparison of reification alternatives.
Give us feedback!
https://kgbook.org 33/150
2/15/23, 5:13 PM Knowledge Graphs
As an alternative to reification, we can rather use higher-arity representations for modelling context.
Taking again the edge Santiago –flight ➛ Arica , Figure 3.9 illustrates three higher-arity representations
of temporal context. First, we can use a named graph (Figure 3.9a) to contain the edge and then de‐
Table of Contents
fine the temporal context on the graph name. Second, we can use a property graph (Figure 3.9b)
where the temporal context is defined as a property on the edge. Third, we can use RDF* [Hartig,
2017] (Figure 3.9c): an extension of RDF that allows edges to be defined as nodes. Amongst these
options, the most flexible is the named graph representation, where we can assign context to multi‐
ple edges at once by placing them in one named graph; for example, we can add more edges to the
named graph of Figure 3.9a that are also valid from 1956. The least flexible option is RDF*, which,
in the absence of an edge id, does not permit different groups of contextual values to be assigned to
an edge; for example, if we add four contextual values to the edge Chile –president ➛ M. Bachelet , to
state that it was valid from 2006 until 2010 and valid from 2014 until 2018, we cannot pair the val‐
ues, but may rather have to create a node to represent different presidencies (in the other models,
we could have used two named graphs or edge ids).
3.3.4 Annotations
Thus far, we have discussed representing context in a graph, but we have not spoken about automat‐
ed mechanisms for reasoning about context; for example, if there are only seasonal summer flights
from Santiago to Arica , we may wish to find other routes from Santiago for winter events taking
place in Arica . While the dates for buses, flights, etc., can be represented directly in the graph, or us‐
ing reification, writing a query to manually intersect the corresponding temporal contexts will be dif‐
ficult. An alternative is to consider annotations that provide mathematical definitions of a contextual
domain and key operations over that domain that can be applied automatically.
Some annotations model a particular contextual domain; for example, Temporal RDF [Gutiérrez et
al., 2007] allows for annotating edges with time intervals, such as Chile –[2006,2010] ➛ M. Bachelet ,
president
while Fuzzy RDF [Straccia, 2009] allows for annotating edges with a degree of truth such as Santiago
– 0.8 ➛ Semi-Arid , indicating that it is more-or-less true – with a degree of 0.8 – that Santiago has a
climate
semi-arid climate.
Give us feedback!
https://kgbook.org 34/150
2/15/23, 5:13 PM Knowledge Graphs
Other forms of annotation are domain-independent; for example, Annotated RDF [Dividino et al.,
2009, Udrea et al., 2010, Zimmermann et al., 2012] allows for representing context modelled as
semi-rings: algebraic structures consisting of domain values (e.g., temporal intervals, fuzzy values,
etc.) and two operators to combine domain values: meet and join.7 We provide
note 7 Theanjoin
example in for annotations is diffe
operator
Figure 3.10, where G is annotated with values from a temporal domain usingalgebra.
sets of integers (1−365
to represent days of the year. For brevity we use intervals, where, e.g., {[150, 152]} denotes the set
{150, 151, 152}. Query Q then asks for flights from Santiago to cities with events; this query will
check and return an annotation reflecting the temporal validity of each answer. To derive these an‐
swers, we require a conjunction of annotations on compatible flight and city edges, using the meet op‐
erator to compute the annotation for which both edges hold. The natural way to define meet here is
as the intersection of sets of days, where, for example, applying meet on the event annotation
{[150, 152]} and the flight annotation {[1, 120], [220, 365]} for Punta Arenas leads to the empty time in‐
terval {}, which may thus lead to the city being filtered from the results (depending on the query
evaluation semantics). However, for Arica , we find two different non-empty intersections:
{[123, 125]} for EID16 and {[276, 279]} for EID17 . Given that we are interested in just the city (a pro‐
jected variable), we can combine the two annotations for Arica using the join operator, returning the
Table of Contents
annotation in which either result holds true. The natural way to define join is as the union of the sets
of days, giving {[123, 125], [276, 279]}.
?city context
Q(G) :
Arica {[123, 125], [276, 279]}
Letting D be a semi-ring imposes that, for any values a, a1 , a2 , a3 in A, the following hold:
(a1 ⊕ a2 ) ⊕ a3 = a1 ⊕ (a2 ⊕ a3 )
(⊥ ⊕ a) = (a ⊕ ⊥) = a
(a1 ⊕ a2 ) = (a2 ⊕ a1 )
(a1 ⊕ a2 ) = (a2 ⊕ a1 )
(a1 ⊗ a2 ) ⊗ a3 = a1 ⊗ (a2 ⊗ a3 )
(⊤ ⊗ a) = (a ⊗ ⊤) = a
a1 ⊗ (a2 ⊕ a3 ) = (a1 ⊗ a2 ) ⊕ (a1 ⊗ a3 )
(⊥ ⊗ a) = (a ⊗ ⊥) = ⊥
(a ⊕ a) = a
(a1 ⊗ a2 ) = (a2 ⊗ a1 )
Give us feedback!
https://kgbook.org 35/150
2/15/23, 5:13 PM Knowledge Graphs
Idempotence induces a partial order: a1 ≤ a2 if and only if a1 ⊕ a2 = a2 . Imposing these condi‐
tions on the annotation domain allow for reasoning and querying to be conducted over the an‐
notation domain in a well-defined manner. Annotated graphs can then be defined in the natural
way:
Example 3.4 — Figure 3.10 exemplifies query answering on a graph annotated with days of the year.
Formally this domain can be defined as follows: A := 2N[1,365] , ⊕ := ∪, ⊗ := ∩, ⊤ := N[1,365] , ⊥ := ∅,
ring.
Table of Contents
Other frameworks have been proposed for modelling and reasoning about context in graphs. A no‐
table example is that of contextual knowledge repositories [Serafini and Homola, 2012], which allow
for assigning individual (sub-)graphs to their own context. Unlike in the case of named graphs, con‐
text is explicitly modelled along one or more dimensions, where each (sub-)graph takes a value for
each dimension. Each dimension is associated with a partial order over its values – e.g., 2020-03-22 ⪯
2020-03 ⪯ 2020 – enabling the selection and combination of sub-graphs that are valid within con‐
texts at different granularities. Schuetz et al. [2021] similarly propose a form of contextual OnLine
Analytic Processing (OLAP), based on a data cube formed by dimensions where each cell contains a
knowledge graph. Operations such as “slice-and-dice” (selecting knowledge according to given dimen‐
sions), as well as “roll-up” (aggregating knowledge at a higher level) are supported. We refer the
reader to the respective papers for more details [Serafini and Homola, 2012, Schuetz et al., 2021].
Chapter 4
Deductive Knowledge
As humans, we can deduce more from the data graph of Figure 2.1 than what the edges explicitly in‐
dicate. We may deduce, for example, that the Ñam festival ( EID15 ) will be located in Santiago, even
though the graph does not contain an edge EID15 –location ➛ Santiago . We may further deduce that the
cities connected by flights must have some airport nearby, even though the graph does not contain
nodes referring to these airports. In these cases, given the data as premises, and some general rules
about the world that we may know a priori, we can use a deductive process to derive new data, al‐
lowing us to know more than what is explicitly given by the data. These types of general premises
and rules, when shared by many people, form part of “commonsense knowledge” [McCarthy, 1990];
conversely, when rather shared by a few experts in an area, they form part of “domain knowledge”,
where, for example, an expert in biology may know that hemocyanin is a protein containing copper
that carries oxygen in the blood of some species of Mollusca and Arthropoda.
Machines, in contrast, do not have a priori access to such deductive faculties; rather they need to be
given formal instructions, in terms of premises and entailment regimes, facilitating similar deductions
to what a human can make. In this way, we will be making more of the meaning (i.e., semantics) of
the graph explicit in a machine-readable format. These entailment regimes formalise the conclusions
Give us feedback!
https://kgbook.org 36/150
2/15/23, 5:13 PM Knowledge Graphs
that logically follow as a consequence of a given set of premises. Once instructed in this manner, ma‐
chines can (often) apply deductions with a precision, efficiency, and scale beyond human perfor‐
mance. These deductions may serve a range of applications, such as improving query answering, (de‐
ductive) classification, finding inconsistencies, etc. As a concrete example involving query answer‐
ing, assume we are interested in knowing the festivals located in Santiago; we may straightforwardly
express such a query as per the graph pattern shown in Figure 4.1. This query returns no results for
the graph in Figure 2.1: there is no node named Festival , and nothing has (directly) the location
Santiago . However, an answer ( Ñam ) could be automatically entailed were we to state that x being a
Food Festival entails that x is a Festival, or that x having venue y in city z entails that x has location
z . How, then, should such entailments be captured? In Section 3.1.1 we already discussed how the
former entailment can be captured with sub-class relations in a semantic schema; the second entail‐
ment, however, requires a more expressive entailment regime than seen thus far.
Table of Contents
In this chapter, we discuss ways in which more complex entailments can be expressed and automat‐
ed. Though we could leverage a number of logical frameworks for these purposes – such as First-
Order Logic, Datalog, Prolog, Answer Set Programming, etc. – we focus on ontologies, which consti‐
tute a formal representation of knowledge that, importantly for us, can be represented as a graph.
We then discuss how these ontologies can be formally defined, how they relate to existing logical
frameworks, and how reasoning can be conducted with respect to such ontologies.
4.1 Ontologies
To enable entailment, we must be precise about what the terms we use mean. Returning to
Figure 2.1, for example, and examining the node EID16 more closely, we may begin to question how
it is modelled, particularly in comparison with EID15 . Both nodes – according to the class hierarchy
of Figure 3.1 – are considered to be events. But what if, for example, we wish to define two pairs of
start and end dates for EID16 corresponding to the different venues? Should we rather consider what
takes place in each venue as a different event? What then if an event has various start and end dates
in a single venue: would these also be considered as one (recurring) event, or many events? These
questions are facets of a more general question: what precisely do we mean by an “event”? Does it hap‐
pen in one contiguous time interval or can it happen many times? Does it happen in one place or can
it happen in multiple? There are no “correct” answers to such questions – we may understand the
term “event” in a variety of ways, and thus the answers are a matter of convention.
Like all conventions, the usefulness of an ontology depends on the level of agreement on what that
ontology defines, how detailed it is, and how broadly and consistently it is adopted. Adoption of an
ontology by the parties involved in one knowledge graph may lead to a consistent use of terms and
Give us feedback!
https://kgbook.org 37/150
2/15/23, 5:13 PM Knowledge Graphs
consistent modelling in that knowledge graph. Agreement over multiple knowledge graphs will, in
turn, enhance the interoperability of those knowledge graphs.
Amongst the most popular ontology languages used in practice are the Web Ontology Language
(OWL) [Hitzler et al., 2012]9, recommended by the W3C and compatible with noteRDF graphs;
9 We couldand the RDF Schema (RDFS) i
include
OWL,
Open Biomedical Ontologies Format (OBOF) [Mungall et al., 2012], used mostly which
in the extends do‐
biomedical its core.
main. Since OWL is the more widely adopted, we focus on its features, though many similar features
are found in both [Mungall et al., 2012]. Before introducing such features, however, we must discuss
how graphs are to be interpreted.
We as humans may interpret the node Santiago in the data graph of Figure 2.1 as referring to the real-
world city that is the capital of Chile. We may further interpret an edge Arica –flight ➛ Santiago as stat‐
ing that there are flights from the city of Arica to this city. We thus interpret the data graph as an‐
other graph – what we here call the domain graph – composed of real-world entities connected by
Table of Contents
real-world relations. The process of interpretation, here, involves mapping the nodes and edges in the
data graph to nodes and edges of the domain graph.
Along these lines, we can abstractly define an interpretation of a data graph as being composed of two
elements: a domain graph, and a mapping from the terms (nodes and edge-labels) of the data graph
to those of the domain graph. The domain graph follows the same model as the data graph; for ex‐
ample, if the data graph is a directed edge-labelled graph, then so too will be the domain graph. For
simplicity, we will speak of directed edge-labelled graphs and refer to the nodes of the domain graph
as entities, and to its edges as relations. Given a data graph and an interpretation, while we denote
nodes in the data graph by Santiago , we will denote the entity it refers to in the domain graph by
Santiago (per the mapping of the given interpretation). Likewise, while we denote an edge by Arica –
flight ➛ Santiago –
, we will denote the relation by Arica flight Santiago (again, per the mapping of the
given interpretation). In this abstract notion of an interpretation, we do not require that Santiago or
Arica be the real-world cities, nor even that the domain graph contain real-world entities and rela‐
tions: an interpretation can have any domain graph and mapping.
Why is such an abstract notion of interpretation useful? The distinction between nodes/edges and
entities/relations becomes important when we define the meaning of ontology features and entail‐
ment. To illustrate this distinction, if we ask whether there is an edge labelled flight between Arica
and Viña del Mar for the data graph in Figure 2.1, the answer is no. However, if we ask if the entities
Arica and Viña del Mar are connected by the relation flight, then the answer depends on what assump‐
tions we make when interpreting the graph. Under the Closed World Assumption (CWA), if we do
not have additional knowledge, then the answer is a definite no – since what is not known is as‐
sumed to be false. Conversely, under the Open World Assumption (OWA), we cannot be certain that
this relation does not exist as this could be part of some knowledge not (yet) described by the graph.
Likewise under the Unique Name Assumption (UNA), the data graph describes at least two flights to
Santiago –
(since Viña del Mar and Arica are assumed to be different entities and, therefore, Arica flight
Santiago –
and Viña del Mar flight Santiago must be different edges). Conversely, under No Unique
Name Assumption (NUNA), we can only say that there is at least one such flight since Viña del Mar
and Arica may be the same entity with two “names”.
These assumptions (or lack thereof) define which interpretations are valid, and which interpretations
satisfy which data graphs. We call an interpretation that satisfies a data graph a model of that data
graph. The UNA forbids interpretations that map two data terms to the same domain term. The
NUNA allows such interpretations. Under the CWA, an interpretation that contains an edge x p – y
in its domain graph can only satisfy a data graph from which we can entail x –p ➛ y . Under the
–
OWA, an interpretation containing the edge x p y can satisfy a data graph not entailing x –p ➛ y
so long it does not explicitly contradict that edge. OWL adopts the NUNA and OWA, which is the
most general case: multiple nodes/edge-labels in the graph may refer to the same entity/relation-
type (per the NUNA), and anything not entailed by the data graph is not assumed to be false as a
consequence (per the OWA).
Give us feedback!
https://kgbook.org 38/150
2/15/23, 5:13 PM Knowledge Graphs
A graph interpretation – or simply interpretation – captures the assumptions under which the se‐
mantics of a graph can be defined. We define interpretations for directed edge-labelled graphs,
though the notion extends naturally to other graph models (assuming the data and domain
graphs follow the same model).
edge-labelled) graph called the domain graph and ⋅I : Con → VΓ ∪ LΓ is a partial mapping from
We denote the domain of the mapping ⋅I by dom(⋅I ). For interpretations under the UNA, the
mapping ⋅I is required to be injective, while with no UNA (NUNA), no such requirement is
necessary.
Interpretations that satisfy a graph are then said to be models of that graph.
Table of Contents
V ∪ L ⊆ dom(⋅I );
for all v ∈ V , it holds that v I ∈ VΓ ;
Beyond our base assumptions, we can associate certain patterns in the data graph with semantic con‐
ditions that define which interpretations satisfy it; for example, we can add a semantic condition to
enforce that if our data graph contains the edge p –subp. of ➛ q , then any edge x p – y in the do‐
main graph of the interpretation must also have a corresponding edge x q – y to satisfy the data
graph. These semantic conditions then form the features of an ontology language. In what follows, to
aid readability, we will introduce the features of OWL using an abstract graphical notation with ab‐
breviated terms. For details of concrete syntaxes, we rather refer to the OWL and OBOF
standards [Hitzler et al., 2012, Mungall et al., 2012]. Likewise we present semantic conditions over
interpretations for each feature in the same graphical format;10 further details
note of
10 these conditions “if and only if” as “i
We abbreviate
will be described later in Section 4.1.3. “if ϕ then ψ ” and “if ψ then ϕ”.
Individuals
In Table 4.1, we list the main features supported by OWL for describing individuals (e.g., Santiago,
EID16), sometimes distinguished from classes and properties. First, we can assert (binary) relations
between individuals using edges such as Santa Lucía –city ➛ Santiago . In the condition column, when
we write x – y z , for example, we refer to the condition that the relation is given in the domain
graph of the interpretation; if so, the interpretation satisfies the axiom. OWL further allows for defin‐
ing relations to explicitly state that two terms refer to the same entity, where, e.g., Región V –same as ➛
Región de Valparaíso states that both refer to the same region (per Section 3.2); or that two terms refer
to different entities, where, e.g., Valparaíso –diff. from ➛ Región de Valparaíso distinguishes the city from
the region of the same name. We may also state that a relation does not hold using negation, which
can be serialised as a graph using a form of reification (see Figure 3.8a).
https://kgbook.org 39/150
2/15/23, 5:13 PM Knowledge Graphs
Negation not x – y z
Same As x1 –same as ➛ x2 x1 = x2
Región V –same as ➛ Región de Valparaíso
Properties
In Section 3.1.1, we already discussed how sub-properties, domains and ranges may be defined for
properties. OWL allows such definitions, and further includes other features, as listed in Table 4.2.
We may define a pair of properties to be equivalent, inverses, or disjoint. We can further define a par‐
Table of Contents
ticular property to denote a transitive, symmetric, asymmetric, reflexive, or irreflexive relation. We can
also define the multiplicity of the relation denoted by properties, based on being functional (many-to-
one) or inverse-functional (one-to-many). We may further define a key for a class, denoting the set of
properties whose values uniquely identify the entities of that class. Without adopting a Unique Name
Assumption (UNA), from these latter three features we may conclude that two or more terms refer to
the same entity. Finally, we can relate a property to a chain (a path expression only allowing con‐
catenation of properties) such that pairs of entities related by the chain are also related by the given
property. Note that for the latter two features in Table 4.2 we require representing a list, denoted
with a vertical notation ⋮ ; while such a list may be serialised as a graph in a number of concrete
ways, OWL uses RDF lists (see Figure 3.7).
x – p y – p z
–
Transitive p –type ➛ Transitive part of –type ➛ Transitive
implies x p z
– p y p – x2
p1
Key –key ➛ implies = –key ➛ long
c ⋮ x1 x2 City lat
pn
Give us feedback!
https://kgbook.org 40/150
2/15/23, 5:13 PM Knowledge Graphs
–chain ➛
q1 x – q1 y1 –… yn−1 – qn z
–chain ➛ location
–
Chain
p ⋮ location part of
qn
implies x p z
Classes
In Section 3.1.1, we discussed how class hierarchies can be modelled using a sub-class relation. OWL
supports sub-classes, and many additional features, for defining and making claims about classes;
these additional features are summarised in Table 4.3. Given a pair of classes, OWL allows for defin‐
ing that they are equivalent, or disjoint. Thereafter, OWL provides a variety of features for defining
novel classes by applying set operators on other classes, or based on conditions that the properties of
its instances satisfy. First, using set operators, one can define a novel class as the complement of an‐
other class, the union or intersection of a list (of arbitrary length) of other classes, or as an enumera‐
tion of all of its instances. Second, by placing restrictions on a particular property p, one can define
classes whose instances are all of the entities that have: some value from a given class on p; all values
Table of Contents
from a given class on p;11 have a specific individual as a value on p (has value);
notehave themselves
11 While as a like flight prop – Domest
something ➛
reflexive value on p (has self); have at least, at most or exactly some number of
bevalues
a moreonnatural
p (cardinali‐
example for All Values, this
ty); and have at least, at most or exactly some number of values on p froming for allclass
a given condition is satisfied when no such
(qualified
known not to have any flights to be a domestic
cardinality). For the latter two cases, in Table 4.3, we use the notation “#{ intersection
a ∣ ϕ}” to count distinct
of such a definition and airport as b
entities satisfying ϕ in the interpretation. These features can then be combined to create more com‐
plex classes, where combining the examples for Intersection and Has Self in Table 4.3 gives the def‐
inition: self-driving taxis are taxis having themselves as a driver.
x – type d1 or
– –
d1
–union ➛ iff or –union ➛ InternationalFlight
–
dn
x type dn
–
d1
Intersection –inter. ➛ iff –inter. ➛ SelfDriving
c ⋮ x type c Taxi
SelfDrivingTaxi
dn
–
x1 Austria
Enumeration –one of ➛ iff –one of ➛
–
there exists a such that
– –
Some Values x type c iff
x p a type d
–
for all a with x – p a
–
All Values x type c iff
it holds that a type d
Cardinality x – type c
Qualified
x – type c
– –
Cardinality
iff #{ a ∣ x p a type d }⋆n
⋆ ∈ {=, ≤, ≥}
Give us feedback!
https://kgbook.org 41/150
2/15/23, 5:13 PM Knowledge Graphs
Other features
OWL supports other language features not previously discussed, including: annotation properties,
which provide metadata about ontologies, such as versioning info; datatype vs. object properties,
which distinguish properties that take datatype values from those that do not; and datatype facets,
which allow for defining new datatypes by applying restrictions to existing datatypes, such as to de‐
fine that places in Chile must have a float between −66.0 and −110.0 as their value for the (datatype)
property latitude. For more details we refer to the OWL 2 standard [Hitzler et al., 2012]. We will fur‐
ther discuss methodologies for the creation of ontologies in Section 6.5.
Each axiom described by the previous tables, when added to a graph, enforces some condition(s) on
the models the graph. If we were to consider only the base condition of the Assertion feature in
Table 4.1, for example, then the models of a graph would be any interpretation such that for every
–
edge x –y ➛ z in the graph, there exists a relation x y z in the model. Given that there may be
other relations in the model (under the OWA), the number of models of any such graph is infinite.
Table of Contents
Furthermore, given that we can map multiple nodes in the graph to one entity in the model (under
the NUNA), any interpretation with (for example) the relation a a – a is a model of any graph so
long as for every edge x –y ➛ z in the graph, it holds that x = y = z = a in the interpretation (in
other words, the interpretation maps everything to a ). As we add axioms with their associated con‐
ditions to the graph, we restrict models for the graph; for example, considering a graph with two
edges – x –y ➛ z and y –type ➛ Irreflexive – the interpretation with a a – a, x = y = … = a is no
longer a model as it breaks the condition for the irreflexive axiom. In this way, we can define a pre‐
cise model-theoretic semantics for graphs based on how the aforementioned ontological features
used in the graph restrict the models of that graph.
We do not restrict the language used to define semantic conditions, but, for example, we can de‐
fine the Has Value semantic condition of Table 4.3 in FOL as:
Here we overload Γ as a ternary predicate to capture the edges of Γ. The other semantic condi‐
tions enumerated in Tables 4.1–4.3 can be defined in a similar way [Schneider and Sutcliffe,
2011].12 This FOL formula defines an if-and-only-if version of the semantic
note condition for these
12 Although Has tables consider axioms o
Value (described in Section 4.1.4). check their image in the domain graph since I
the axioms are reflected in the domain graph of
the definitions considerably.
4.1.3 Entailment
The conditions listed in the previous tables give rise to entailments, where, for example, in reference
to the Symmetric feature of Table 4.2, the definition nearby –type ➛ Symmetric and edge Santiago –
nearby ➛ Santiago Airport entail the edge Santiago Airport –nearby ➛ Santiago according to the condition giv‐
en for that feature. We now describe how these conditions lead to entailments.
We say that one graph entails another if and only if any model of the former graph is also a model of
the latter graph. Intuitively this means that the latter graph says nothing new over the former graph
and thus holds as a logical consequence of the former graph. For example, consider the graph
Santiago –type ➛ City –subc. of ➛ Place and the graph Santiago –type ➛ Place . All models of the latter must
–
have that Santiago type Place , –
but so must all models of the former, which must have Santiago typeGive us feedback!
https://kgbook.org 42/150
2/15/23, 5:13 PM Knowledge Graphs
–
City subc. of Place and further must satisfy the condition for Sub-class, which requires that
–
Santiago type Place also hold. Hence we conclude that any model of the former graph must be a
model of the latter graph, or, in other words, the former graph entails the latter graph.
tions, we say that G1 entails G2 under Φ – denoted G1 ⊨Φ G2 – if and only if any model of G1
–
nearby type
Condition holds in domain graph – the graphs entail each other: both graphs while nearbytoinv.
give rise theofsame
nearby is a model of the se
condition, which is translated back into all possible axioms that describe it. ther
Hence if-and-only-if
graph entails these‐
other.
mantics allows for entailing more axioms in the ontology language than if–then semantics. OWL gen‐
erally applies an if-and-only-if semantics in order to enable richer entailments [Hitzler et al., 2012].
4.2 Reasoning
Unfortunately, given two graphs, deciding if the first entails the second – per the notion of entail‐
ment we have defined and for all of the ontological features listed in Tables 4.1–4.3 – is undecidable:
no (finite) algorithm for such entailment can exist that halts on all inputs with the correct
true/false answer [Hitzler et al., 2010]. However, we can provide practical reasoning algorithms
for ontologies that (1) halt on any pair of input ontologies but may miss entailments, returning
false instead of true in some cases, (2) always halt with the correct answer but only accept input
ontologies with restricted features, or (3) only return correct answers for any pair of input ontologies
but may never halt on certain inputs. Though option (3) has been explored using, e.g., theorem
provers for First Order Logic (FOL) [Schneider and Sutcliffe, 2011], options (1) and (2) are more
commonly pursued using rules and/or Description Logics. Option (1) generally allows for more effi‐
cient and scalable reasoning algorithms and is useful where data are incomplete and having some en‐
tailments is valuable. Option (2) may be a better choice in domains – such as medical ontologies –
where missing entailments may have undesirable outcomes.
4.2.1 Rules
A straightforward way to provide automated access to the knowledge that can be deduced through
(ontological or other forms of) entailments is through inference rules (or simply rules) encoding if–
then-style consequences. A rule is composed of a body (if) and a head (then). Both the body and
head are given as graph patterns. A rule indicates that if we can replace the variables of the body
with terms from the data graph and form a sub-graph of a given data graph, then using the same re‐
placement of variables in the head will yield a valid entailment. The head must typically use a subset
of the variables appearing in the body to ensure that the conclusion leaves no variables unreplaced.
Rules of this form correspond to (positive) Datalog [Ceri et al., 1989] in Databases, Horn
clauses [Lloyd, 1984] in Logic Programming, etc. Give us feedback!
https://kgbook.org 43/150
2/15/23, 5:13 PM Knowledge Graphs
Rules can capture entailments under ontological conditions. In Table 4.4, we list some example rules
for sub-class, sub-property, domain and range features [Muñoz et al., 2009]; these rules may be con‐
sidered incomplete, not capturing, for example, that every class is a sub-class of itself, that every
property is a sub-property of itself, etc. A more comprehensive set of rules for the OWL features of
Tables 4.1–4.3 have been defined as OWL 2 RL/RDF [Motik et al., 2012]; these rules are likewise in‐
complete as such rules cannot fully capture negation (e.g., Complement), existentials (e.g., Some
Values), universals (e.g., All Values), or counting (e.g., Cardinality and Qualified Cardinality).
Other rule languages have, however, been proposed to support additional such features, including
existentials (see, e.g., Datalog± [Bellomarini et al., 2018]), disjunction (see, e.g., Disjunctive
Datalog [Rudolph et al., 2008]), etc.
Table 4.4: Example rules for sub-class, sub-property, domain, and range
features
Domain ⇒ ?x –type ➛ ?c
Range ⇒ ?y –type ➛ ?c
Rules can be leveraged for reasoning in a number of ways. Materialisation refers to the idea of apply‐
ing rules recursively to a graph, adding the conclusions generated back to the graph until a fixpoint
is reached and nothing more can be added. The materialised graph can then be treated as any other
graph. Although the efficiency and scalability of materialisation can be enhanced through optimisa‐
tions like Rete networks [Forgy, 1982], or using distributed frameworks like MapReduce [Urbani et
al., 2012], depending on the rules and the data, the materialised graph may become unfeasibly large
to manage. Another strategy is to use rules for query rewriting, which given a query, will automatical‐
ly extend the query in order to find solutions entailed by a set of rules; for example, taking the
schema graph in Figure 3.2 and the rules in Table 4.4, the (sub-)pattern ?x –type ➛ Event in a given
input query would be rewritten to the following disjunctive pattern evaluated on the original graph:
Figure 4.2 provides a more complete example of an ontology that is used to rewrite the query of
Figure 4.1; if evaluated over the graph of Figure 2.1, Ñam will be returned as a solution. However,
not all of the aforementioned features of OWL can be supported in this manner. The OWL 2 QL pro‐
file [Motik et al., 2012] is a subset of OWL designed specifically for query rewriting of this
form [Artale et al., 2009].
O:
Q(O) : ( ∪ ∪ )
⋈( ∪ )
⋈
Figure 4.2: Query rewriting example for the query Q of Figure 4.1
While rules can be used to (partially) capture ontological entailments, they can also be defined inde‐
pendently of an ontology language, capturing entailments for a given domain. In fact, some rules –
such as the following – cannot be captured by the ontology features previously seen, as they do not
support ways to infer relations from cyclical graph patterns (for computability reasons):
⇒ ?x –domestic flight ➛ ?y
Give us feedback!
https://kgbook.org 44/150
2/15/23, 5:13 PM Knowledge Graphs
Various languages allow for expressing rules over graphs – independently or alongside of an ontol‐
ogy language – including: Notation3 (N3) [Berners-Lee and Connolly, 2011], Rule Interchange
Format (RIF) [Kifer and Boley, 2013], Semantic Web Rule Language (SWRL) [Horrocks et al., 2004],
and SPARQL Inferencing Notation (SPIN) [Knublauch et al., 2011], amongst others.
Given a graph pattern Q – be it a directed edge-labelled graph pattern per Definition 2.5 or a
property graph pattern per Definition 2.6 – recall that Var(Q) denotes the variables appearing
in Q. We now define rules for graphs.
This definition of a rule applies for directed edge-labelled graphs and property graphs by consid‐
ering the corresponding type of graph pattern. The head is considered to be a conjunction of
edges. Given a graph G, a rule is applied by computing the mappings from the body to the graph
Table of Contents
and then using those mappings to substitute the variables in H . The restriction Var(H) ⊆ B en‐
sures that the results of this substitution is a graph, with no variables in H left unsubstituted.
Given a set of rules R := {R1 , … , Rn } and a knowledge graph G, towards defining the set of in‐
ferences given by the rules over the graph, we denote by R(G) := ⋃R∈R R(G) the union of the
application of all rules of R over G, and we denote by R+ (G) := R(G) ∪ G the extension of G
with respect to the application of R. Finally, we denote by Rk (G) (for k ∈ N+ ) the recursive ap‐
plication of R+ (G), where R1 (G) := R+ (G), and Ri+1 (G) := R+ (Ri (G)). We are now ready to
define the least model, which captures the inferences possible for R over G.
At some point Rk (G) = Rk +1 (G): the rule applications reach a fixpoint and we have the least
′ ′
model. Once the least model R∗ (G) is computed, the entailed data can be treated as any other
data.
Rules can support graph entailments of the form G1 ⊨Φ G2 . We say that a set of rules R is cor‐
rect for Φ if, for any graph G, G ⊨Φ R∗ (G). We say that R is complete for Φ if, for any graph G,
incomplete set of rules for the semantic conditions of the RDFS standard [Brickley and Guha,
2014].
Alternatively, rather than supporting ontology-based graph entailments, rules can be directly
specified in a rule language such as Notation3 (N3) [Berners-Lee and Connolly, 2011], Rule
Interchange Format (RIF) [Kifer and Boley, 2013], Semantic Web Rule Language
(SWRL) [Horrocks et al., 2004], or SPARQL Inferencing Notation (SPIN) [Knublauch et al.,
2011]. Languages such as SPIN represent rules as graphs, allowing the rules of a knowledge
graph to be embedded in the data graph. Taking advantage of this fact, we can then consider a
form of graph entailment G1 ∪ γ(R) ⊨Φ G2 , where by γ(R) we denote the graph representation
of rules R. If the set of rules R is correct and complete for Φ, we may simply write G1 ∪ γ(R) ⊨
G2 , indicating that Φ captures the same semantics for γ(R) as applying the rules in R. Rules
https://kgbook.org 45/150
2/15/23, 5:13 PM Knowledge Graphs
Description Logics (DLs) were initially introduced as a way to formalise the meaning of frames and
semantic networks . Since semantic networks are an early version of knowledge graphs, and DLs have
heavily influenced the Web Ontology Language, DLs thus hold an important place in the logical for‐
malisation of knowledge graphs. DLs form a family of logics rather than a particular logic. Initially,
DLs were restricted fragments of FOL that permit decidable reasoning tasks, such as entailment
checking [Baader et al., 2017]. Different DLs strike different balances between expressive power and
computational complexity of reasoning. DLs were later extended with features beyond FOL that are
useful in the context of modelling graph data, such as transitive closure, datatypes, etc.
DLs are based on three types of elements: individuals, such as Santiago; classes (aka concepts) such
as City; and properties (aka roles) such as flight. DLs then allow for making claims, known as ax‐
ioms, about these elements. Assertional axioms can be either unary class relations on individuals, such
asCity(Santiago), or binary property relations on individuals, such as flight(Santiago,Ar‐
ica). Such axioms form the Assertional Box (A-Box). DLs further introduce logical symbols to allow
for defining class axioms (forming the Terminology Box, or T-Box for short), and property axioms
(forming the Role Box, R-Box); for example, the class axiom City ⊑ Place states that the former
Table of Contents
class is a sub-class of the latter one, while the property axiom flight ⊑ connectsTo states that
the former property is a sub-property of the latter one. DLs may then introduce a rich set of logical
symbols, not only for defining class and property axioms, but also defining new classes based on ex‐
isting terms; as an example of the latter, we can define a class ∃nearby.Airport as the class of in‐
dividuals that have some airport nearby. Noting that the symbol ⊤ is used in DLs to denote the class
of all individuals, we can then add a class axiom ∃flight.⊤ ⊑ ∃nearby.Airport to state that indi‐
viduals with an outgoing flight must have some airport nearby. Noting that the symbol ⊔ can be
used in DL to define that a class is the union of other classes, we can further define, for example, that
Airport ⊑ DomesticAirport ⊔ InternationalAirport, i.e., that an airport is either a domes‐
tic airport or an international airport (or both).
The similarities between DL features and the OWL features seen previously are not coincidental: the
OWL standard was heavily influenced by DLs, where, for example, the OWL 2 DL language is a frag‐
ment of OWL restricted so that entailment becomes decidable, where the restrictions are inspired by
those defined for DLs. To exemplify a restriction,
DomesticAirport ⊑ = 1 destination ∘ country.⊤ defines in DL syntax that domestic airports
have flights destined to precisely one country (where p ∘ q denotes a chain of properties). However,
counting chains (in this case with = 1 destination ∘ country) is often disallowed in DLs to ensure
decidability.
Expressive DLs support complex entailments involving existentials, universals, counting, etc. A com‐
mon strategy for deciding such entailments is to reduce entailment to satisfiability, which decides if
an ontology is consistent or not [Horrocks and Patel-Schneider, 2004].15 Thereafter methods
note 15 G such
entails G as only if G ∪ not(G′ ) is n
′ if and
tableau can be used to check satisfiability, cautiously constructing models by completing them along
similar lines to the materialisation strategy previously described, but additionally branching models
in the case of disjunction, introducing new elements to represent existentials, etc. If any model is
successfully “completed”, the process concludes that the original definitions are satisfiable (see,
e.g., [Motik et al., 2009]). Due to their prohibitive computational complexity [Motik et al., 2012] –
where for example, disjunction may lead to an exponential number of branching possibilities – such
reasoning strategies are not typically applied in the case of large-scale data, though they may be use‐
ful when modelling complex domains for knowledge graphs.
Table 4.5 provides definitions for all of the constructs typically found in Description Logics. The
syntax column denotes how the construct is expressed in DL. The semantics column defines the Give us feedback!
https://kgbook.org 46/150
2/15/23, 5:13 PM Knowledge Graphs
meaning of axioms using interpretations, which are defined in a slightly different way to those
seen previously for graphs.
An interpretation I satisfies a knowledge-base K if and only if, for all of the syntactic axioms in
K, the corresponding semantic conditions in Table 4.5 hold for I . In this case, we call I a model
of K.
⚓ , 🏔 , ✈};
Δ I := {
AricaI := ⚓ , SantiagoI := 🏔 , AricaAirportI := ✈;
CityI := {⚓ , 🏔 }, AirportI := {✈};
flightI := {(⚓ , 🏔 )}, connectsToI := {(⚓ , 🏔 )}, sellsI := {(✈, ☕ )}.
The interpretation I is not a model of K since it does not have that ⚓ is nearby some Airport, nor
that ⚓ and 🏔 are in the class Place. However, if we extend the interpretation I with the following:
PlaceI := {⚓ , 🏔 };
nearby := {(⚓ , ✈)}.
Now I is a model of K. Note that although K does not imply that sells(AricaAirport,coffee)
while I indicates that ✈ does indeed sell ☕ , I is still a model of K since K is not assumed to be a com‐
plete description, per the OWA.
Finally, the notion of a model gives rise to the notion of entailment, which tells us which knowl‐
edge bases hold as a logical consequence of which others.
Example 4.2 — Let K1 denote the knowledge base K from the Example 4.1, and define a second
Unfortunately, the problem of deciding entailment for knowledge bases expressed in the DL
composed of the unrestricted use of all of the axioms of Table 4.5 is undecidable since we could
reduce instances of the Halting Problem to such entailment. Hence DLs in practice restrict use of
the features listed in Table 4.5. Different DLs apply different restrictions, implying different
trade-offs for expressivity and the complexity of entailment. Most DLs are founded on one of the
following base DLs (we use indentation to denote derivation):
Give us feedback!
https://kgbook.org 47/150
2/15/23, 5:13 PM Knowledge Graphs
[ALC ] (Attributive Language with C omplement} [Schmidt-Schauß and Smolka, 1991]),
supports atomic classes, the top and bottom classes, class intersection, class union, class
negation, universal restrictions and existential restrictions. Relation and class assertions
are also supported.
[S ] extends ALC with transitive closure.
We use “(limited)” to indicate that such features are often only allowed under certain restric‐
tions to ensure decidability; for example, complex relations (chains) typically cannot be com‐
bined with cardinality restrictions. DLs are then typically named per the following scheme,
where [a∣b] denotes an alternative between a and b and [c][d] denotes a concatenation cd:
[ALC∣S][H∣R][O][I][F∣N ∣Q]
Examples include ALCO, ALCHI , SHIF , SROIQ, etc. These languages often apply additional
restrictions on class and property axioms to ensure decidability, which we do not discuss here.
For further details on DLs, we refer to the recent book by Baader et al. [2017].
As mentioned in the body of the survey, DLs have been very influential in the definition of OWL,
where the OWL 2 DL fragment (roughly) corresponds to the DL SROIQ. For example, the ax‐
iom venue –domain ➛ Event in OWL can be translated to ∃venue.⊤ ⊑ Event, meaning that the
class of individuals with some value for venue (in any class) is a sub-class of the class Event.
We leave other translations from the OWL axioms of Tables 4.1–4.3 to note DL as
16 an exercise. 16
Though not previously mentioned, OW
and Nothing
Note, however, that axioms like sub-taxon of –subp. of ➛ subc. of – which given that correspond
a graph such as to ⊤ and ⊥, respe
Fred –type ➛ Homo sapiens –sub-taxon of ➛ Hominini entails the edge Fred –type ➛ Hominini – cannot be
expressed in DL: “subTaxonOf ⊑ ⊑” is not syntactically valid. Hence only a subset of graphs
can be translated into well-formed DL ontologies; we refer to the OWL standard for
details [Hitzler et al., 2012].
Table 4.5: Description Logic semantics (such that x, y, z, aI , aI1 , … aIn , bI are in ΔI )
https://kgbook.org 48/150
2/15/23, 5:13 PM Knowledge Graphs
Name Syntax Semantics (⋅I )
Universal Relation U ΔI × ΔI
Relation Axioms (R-Box)
Relation Inclusion R⊑S RI ⊆ S I
Complex Relation Inclusion R1 ∘ ... ∘ Rn ⊑ S
R1I ∘ ... ∘ RnI ⊆ S I
Chapter 5
Inductive Knowledge
While deductive knowledge is characterised by precise logical consequences, inductively acquiring
knowledge involves generalising patterns from a given set of input observations, which can then be
used to generate novel but potentially imprecise predictions. For example, from a large data graph
with geographical and flight information, we may observe the pattern that almost all capital cities of
countries have international airports serving them, and hence predict that if Santiago is a capital
city, it likely has an international airport serving it; however, the predictions drawn from this pattern
do not hold for certain, where (e.g.) Vaduz, the capital city of Liechtenstein, has no (international)
airport serving it. Hence predictions will often be associated with a level of confidence; for example,
we may say that a capital has an international airport in 187
195 of cases, offering a confidence of 0.959
for predictions made with that pattern. We then refer to knowledge acquired inductively as inductive
knowledge, which includes both the models used to encode patterns, as well as the predictions made
by those models. Though fallible, inductive knowledge can be highly valuable.
Figure 5.1: Conceptual overview of popular inductive techniques for knowledge graphs in terms of type of
representation generated (Numeric/Symbolic) and type of paradigm used (Unsupervised/Self-
supervised/Supervised)
In Figure 5.1 we provide an overview of the inductive techniques typically applied to knowledge
graphs. In the case of unsupervised methods, there is a rich body of work on graph analytics, which
uses well-known functions/algorithms to detect communities or clusters, find central nodes and
Give us feedback!
https://kgbook.org 49/150
2/15/23, 5:13 PM Knowledge Graphs
edges, etc., in a graph. Alternatively, knowledge graph embeddings can use self-supervision to learn a
low-dimensional numeric model of a knowledge graph that (typically) maps input edges to an output
plausibility score indicating the likelihood of the edge being true. The structure of graphs can also be
directly leveraged for supervised learning, as explored in the context of graph neural networks.
Finally, while the aforementioned techniques learn numerical models, symbolic learning can learn
symbolic models – i.e., logical formulae in the form of rules or axioms – from a graph in a self-super‐
vised manner. We now discuss each of the aforementioned techniques in turn.
Analytics is the process of discovering, interpreting, and communicating meaningful patterns inher‐
ent to (typically large) data collections. Graph analytics is then the application of analytical process‐
es to (typically large) graph data. The nature of graphs naturally lends itself to certain types of ana‐
lytics that derive conclusions about nodes and edges based on the topology of the graph, i.e., how the
nodes of the graph are connected. Graph analytics draws upon techniques from related areas, such as
Table of Contents
graph theory and network analysis, which have been used to study graphs representing social net‐
works, the Web, internet routing, transport networks, ecosystems, protein–protein interactions, lin‐
guistic cooccurrences, and more besides [Estrada, 2011].
Returning to the domain of our running example, the tourism board could use graph analytics to ex‐
tract knowledge about, for instance: key transport hubs that serve many tourist attractions (centrali‐
ty); groupings of attractions visited by the same tourists (community detection); attractions that may
become unreachable in the event of strikes or other route failures (connectivity), or pairs of attrac‐
tions that are similar to each other (node similarity). Given that such analytics will require a com‐
plex, large-scale graph, for the purposes of illustration, in Figure 5.2 we present a more concise ex‐
ample of some transportation connections in Chile directed towards popular tourist destinations. We
first introduce a selection of key techniques that can be applied for graph analytics. We then discuss
frameworks and languages that can be used to compute such analytics in practice. Given that many
traditional graph algorithms are defined for unlabelled graphs, we then describe ways in which ana‐
lytics can be applied over directed edge-labelled graphs. Finally we discuss the potential connections
between graph analytics and querying and reasoning.
5.1.1 Techniques
A wide variety of techniques can be applied for graph analytics. In the following we will enumerate
some of the main techniques – as recognised, for example, by the survey of Iosup et al. [2016] – that
can be invoked in this setting.
Centrality: aims to identify the most important (aka central) nodes or edges of a graph. Specific
node centrality measures include degree, betweenness, closeness, Eigenvector, PageRank, HITS,
Katz, amongst others. Betweenness centrality can also be applied to edges. For example, a node
centrality measure might predict the transport hubs in Figure 5.2, while edge centrality might
predict traffic by finding connections on which many shortest routes depend.
Community detection: aims to identify communities in a graph, i.e., sub-graphs that are more
densely connected internally than to the rest of the graph. Community detection algorithms,
Give us feedback!
https://kgbook.org 50/150
2/15/23, 5:13 PM Knowledge Graphs
such as minimum-cut algorithms, label propagation, Louvain modularity, amongst others, can dis‐
cover such communities. Community detection applied to Figure 5.2 may, for example, detect a
community to the left (the north of Chile), to the right (the south of Chile), and perhaps also
the centre (Chilean cities with airports).
Connectivity: aims to estimate how well-connected the graph is, revealing, for instance, the re‐
silience and (un)reachability of elements of the graph. Specific techniques include measuring
graph density or k -connectivity, detecting strongly connected components and weakly connected
components, computing spanning trees or minimum cuts, etc. In the context of Figure 5.2, such
analysis may tell us that routes to Grey Glacier , Osorno Volcano and Piedras Rojas are the most
“brittle”, becoming disconnected if one of two bus routes fails.
Node (or vertex) similarity: aims to find nodes that are similar to other nodes by virtue of how
they are connected within their neighbourhood. Node similarity metrics may be computed us‐
ing structural equivalence, random walks, diffusion kernels, etc. These methods provide an under‐
standing of what connects nodes, and, thereafter, in what ways they are similar. In the context
of Figure 5.2, such analysis may tell us that Calama and Arica are similar nodes based on both
having return flights to Santiago and return buses to San Pedro .
Table of Contents
While the previous techniques accept a graph alone as input,17 other forms note
of graph analytics
17 Node maycan be run over an entir
similarity
further accept a node, a pair of nodes, etc., along with the graph. for each node, or can also be run for a specific n
are also measures for graph similarity (based
Path finding: aims to find paths in a graph, typically between pairs ofBobed,
nodes2018]) thatinput.
given as accept multiple graphs as inp
Various technical definitions exist that restrict the set of valid paths between such nodes, in‐
cluding simple paths that do not visit the same node twice, shortest paths that visit the fewest
number of edges, or – as previously discussed in Section 4.3.2 – regular path queries that re‐
strict the labels of edges that can be traversed by the path according to a regular
expression [Angles et al., 2017]. We could use such algorithms to find, for example, the short‐
est path(s) in Figure 5.2 from Torres del Paine to Moon Valley .
Most of the aforementioned techniques for graph analytics were originally proposed and studied for
simple graphs or directed graphs without edge labels. We will discuss their application to more com‐
plex graph models – and how they can be combined with other techniques such as reasoning and
querying – later in Section 5.1.3.
5.1.2 Frameworks
Various frameworks have been proposed for large-scale graph analytics, often in a distributed (clus‐
ter) setting. Amongst these we can mention Apache Spark (GraphX) [Xin et al., 2013a, Dave et al.,
2016], GraphLab [Low et al., 2012], Pregel [Malewicz et al., 2010], Signal–Collect [Stutz et al.,
2016], Shark [Xin et al., 2013b], etc. These graph parallel frameworks apply a systolic abstrac‐
tion [Kung, 1982] based on a directed graph, where nodes are seen as processors that can send mes‐
sages to other nodes along edges. Computation is then iterative, where in each iteration, each node
reads messages received through inward edges (and possibly its own previous state), performs a
computation, and then sends messages through outward edges based on the result. These frame‐
works then define the systolic computational abstraction on top of the data graph being processed:
nodes and edges in the data graph become nodes and edges in the systolic graph.
To take an example, assume we wish to compute the places that are most (or least) easily reached by
the routes shown in the graph of Figure 5.2. A good way to measure this is using centrality, where
we choose PageRank [Page et al., 1999], which computes the probability of a tourist randomly fol‐
lowing the routes shown in the graph being at a particular place after a given number of “hops”. We
can implement PageRank on large graphs using a graph parallel framework. In Figure 5.3, we pro‐
vide an example of an iteration of PageRank for an illustrative sub-graph of Figure 5.2. The nodes
are initialised with a score of 1
∣V ∣
= 16 , where we assume the tourist to have an equal chance of start‐
ing at any point. In the message phase (Msg), each node v passes a score of dRi (v)
of node v in iteration i (the probability of the tourist being at node v after i hops), and by ∣E(v)∣ the
Give us feedback!
https://kgbook.org 51/150
2/15/23, 5:13 PM Knowledge Graphs
number of outgoing edges of v . The aggregation phase (Agg) for v then sums all incoming messages
received along with its constant share of the damping factor ( 1−d
∣V ∣ ) to compute Ri+1 (v). We then pro‐
ceed to the message phase of the next iteration, continuing until some termination criterion is
reached (e.g., iteration count or residual threshold, etc.) and final scores are output.
While the given example is for PageRank, the systolic abstraction is general enough to support a
wide variety of graph analytics, including those previously mentioned. An algorithm in this frame‐
work consists of the functions to compute message values in the message phase (Msg), and to accumu‐
late the messages in the aggregation phase (Agg). The framework will take care of distribution, mes‐
sage passing, fault tolerance, etc. However, such frameworks – based on message passing between
neighbours – have limitations: not all types of analytics can be expressed in such frameworks [Xu et
al., 2019].18 Hence frameworks may allow additional features, such as a global
note step that performs
18 Formally,
Xu eta al. [2019] have shown
the (incomplete)
global computation on all nodes, making the result available to each node [Malewicz Weisfeiler–Lehman (WL) gra
et al., 2010];
graphs. This test involves nodes recursively has
or a mutation step that allows for adding or removing nodes and edges during processing
received from[Malewicz
neighbours, and passing these has
et al., 2010].
Before defining a graph parallel framework, in the interest of generality, we first define a direct‐
ed graph labelled with feature vectors, which captures the type of input that such a framework
can accept, with vectors on both nodes and edges.
Example 5.1 — We define a directed vector-labelled graph in preparation for later computing
PageRank using a graph parallel framework. Let G = (V , E, L) denote a directed edge-labelled graph.
Let ∣E(u)∣ denote the outdegree of node u ∈ V . We then initialise a directed vector-labelled graph
G′ = (V , E ′ , F , λ) such that E ′ = {(x, z) ∣ ∃y : (x, y, z) ∈ E}, and for all u ∈ V , we define λ(u) :=
⎡ ⎤
1
∣V ∣
∣E (u)∣ , and λ(u, v) := [ ], with F := {λ(u) ∣ u ∈ V } ∪ {λ(u, v) ∣ (u, v) ∈ E ′ }, assigning each node
′
⎣ ∣V ∣ ⎦
a vector containing its initial PageRank score, the outdegree of the node, and the number of nodes in
the graph. Conversely, edge-vectors are not used in this case.
We now define a graph parallel framework, where we use {{⋅}} to denote a multiset, 2S→N to de‐
note the set of all multisets containing (only) elements from the set S , and Ra to denote the set
of all vectors of dimension a (i.e., the set of all vectors containing a real-valued elements).
Give us feedback!
https://kgbook.org 52/150
2/15/23, 5:13 PM Knowledge Graphs
Msg: Ra × Rb → Rc
c
Agg: Ra × 2R →N
→ Ra
a
End: 2R →N
→ {true, false}
The function Msg defines what message (i.e., vector) must be passed from a node to a neigh‐
bouring node along a particular edge, given the current feature vectors of the node and the
edge; the function Agg is used to compute a new feature vector for a node, given its previous
feature vector and incoming messages; the function End defines a condition for termination of
vector computation. The integers a, b and c denote the dimensions of node feature vectors, edge
feature vectors, and message vectors, respectively; we assume that a and b correspond with the
dimensions of input feature vectors for nodes and edges. Given a GPF G = (Msg, Agg, End), a di‐
Table of Contents
rected vector-labelled graph G = (V , E, F , λ), and a node u ∈ V , we define the output vector as‐
signed to node u in G by G (written G(G, u)) as follows. First let n(0)
u := λ(u). For all i ≥ 1, let:
u := u
where Mu is the multiset of messages received by node u during iteration i, and nu is the state
(i) (i)
(vector) of node u at the end of iteration i. If j is the smallest integer for which End({{nu ∣ u ∈
(j)
This particular definition assumes that vectors are dynamically computed for nodes, and that
messages are passed only to outgoing neighbours, but the definitions can be readily adapted to
consider dynamic vectors for edges, or messages being passed to incoming neighbours, etc. We
now provide an example instantiating a GPF to compute PageRank over a directed graph.
Example 5.2 — We take as input the directed vector labelled graph G′ = (V , E, F , λ) from
Example 5.1 for a PageRank GPF. First we define the messages passed from u to v :
where d denotes PageRank’s constant dampening factor (typically d := 0.85) and (nv )k denotes the k th
element of the nv vector. In other words, v will pass to u its PageRank score multiplied by the dampen‐
ing factor and divided by its out-degree (we do not require λ(v, u) in this particular example). Next we
define the function for u to aggregate the messages it receives from other nodes:
Agg(nu , Mu ) :=
⎣ ⎦
(nu )2
(nu )3
Here, we sum the scores received from other nodes along with its share of rank from the dampening
factor, copying over the node’s degree and the total number of nodes for future use. Finally, there are a
number of ways that we could define the termination condition; here we simply define:
(i)
End({{nu ∣ u ∈ V }}) := (i ≥ z)
We may note in this example that the total number of nodes is duplicated in the vector for each
node of the graph. Part of the benefit of GPFs is that only local information in the neighbour‐
hood of the node is required for each computation step. In practice, such frameworks may allow
additional features, such as global computation steps whose results are made available to all
nodes [Malewicz et al., 2010], operations that dynamically modify the graph [Malewicz et al.,
2010], etc.
Give us feedback!
https://kgbook.org 53/150
2/15/23, 5:13 PM Knowledge Graphs
5.1.3 Analytics on data graphs
As aforementioned, most analytics presented thus far are, in their “native” form, applicable for undi‐
rected or directed graphs without the edge metadata – i.e., edge labels or property–value pairs – typi‐
cal of graph data models.19 A number of strategies can be applied to make data
notegraphs
19 Wesubject
remarktothat
an‐in the case of property g
alytics of this form: be converted by mapping values to nodes and p
label.
Projection involves simply “projecting” an undirected or directed graph by optionally selecting
a sub-graph from the data graph from which all edge meta-data are dropped; for example, the
graph of Figure 5.3 may be the result of extracting the sub-graph induced by the edge labels
bus and flight from a larger data graph, where the labels are then dropped to create a directed
graph.
Weighting involves converting edge meta-data into numerical values according to some func‐
tion. Many of the aforementioned techniques are easily adapted to the case of weighted (direct‐
ed) graphs; for example, we could consider weights on the graph of Figure 5.3 denoting trip
duration (or price, traffic, etc.), and then compute the shortest (weighted) paths considering
Table of Contents
time by adding the duration of each leg of the respective journey.20 In note
the 20
absence
Otherofforms
external
of analytics are possible if w
weights, we may rather map edge labels to weights, assigning the same ample, if we to
weight annotated
all flightthe graph of Figure 5.3 w
one place to the next, we could leverage Marko
edges, to all bus edges, etc., based on some criteria. reducibility, periodicity, transience, recurrence
Transformation involves transforming the graph to a lower arity model.routes [Dynkin, 1965].
A transformation may
be lossy, meaning that the original graph cannot be recovered; or lossless, meaning that the
original graph can be recovered. Figure 5.4 provides an example of a lossy and lossless trans‐
formation from a directed edge-labelled graph to directed graphs. In the lossy transformation,
we cannot tell, for example, if the original graph contained the edge Iquique –flight ➛ Santiago , or
rather the edge Iquique –flight ➛ Arica , or both. The lossless transformation must introduce new
nodes (similar to reification) to maintain information about directed labelled edges. Both trans‐
formed graphs further attempt to preserve the directionality of the original graph.
Customisation involves changing the analytical procedure to incorporate edge meta-data, such
as was the case for path finding based on path expressions. Other examples might include
structural measures for node similarity that not only consider common neighbours, but also
common neighbours connected by edges with the same label, or aggregate centrality measures
that capture the importance of edges grouped by label, etc.
The results of an analytical process may change drastically depending on which of the previous
strategies are chosen to prepare the graph for analysis. The choice of strategy may be a non-trivial
one to make a priori and may require empirical validation. More study is required to more generally
understand the effects of such strategies on the results of different analytical techniques over differ‐
ent graph models.
As discussed in Section 2.2, various languages for querying graphs have been proposed down
through the years [Angles et al., 2017]. One may consider a variety of ways in which query lan‐
guages and analytics can complement each other. First, we may consider using query languages to
Give us feedback!
https://kgbook.org 54/150
2/15/23, 5:13 PM Knowledge Graphs
project or transform a graph suitable for a particular analytical task, such as to extract the graph of
Figure 5.2 from a larger data graph. Query languages such as SPARQL [Harris et al., 2013],
Cypher [Francis et al., 2018], and G-CORE [Angles et al., 2018] allow for outputting graphs, where
such queries can be used to select sub-graphs for analysis. These languages can also express some
limited (non-recursive) analytics, where aggregations can be used to compute degree centrality, for
example; they may also have some built-in analytical support, where, for example, Cypher [Francis
et al., 2018] allows for finding shortest paths. In the other direction, analytics can contribute to the
querying process in terms of optimisations, where, for example, analysis of connectivity may suggest
how to better distribute a large data graph over multiple machines for querying using, e.g., minimum
cuts [Akhter et al., 2018, Janke et al., 2018]. Analytics have also been used to rank query results
over large graphs [Wagner et al., 2012, Fan et al., 2013], selecting the most important results for
presentation to the user.
In some use-cases we may further wish to interleave querying and analytical processes. For example,
from the full data graph collected by the tourist board, consider an upcoming airline strike where the
board wishes to find the events during the strike with venues in cities unreachable from Santiago by public
Table of Contents
transport due to the strike. Hypothetically, we could use a query to extract the transport network ex‐
cluding the airline’s routes (assuming, per Figure 2.3a that the airline information is available), use
analytics to extract the strongly connected component containing Santiago, and finally use a query to
find events in cities not in the Santiago component on the given dates.21 While
note one couldasolve
21 Such task this
could not be solved in a si
task using an imperative language such as Gremlin [Rodriguez, 2015], GraphX such
[Xinexpressions would
et al., 2013a], or not be capable of filteri
lar airline.
R [The R Foundation, 1992], more declarative languages are also being explored to express such
tasks, with proposals including the extension of graph query languages with recursive
capabilities [Bischof et al., 2012, Reutter et al., 2015, Hogan et al., 2020],22note
combining linearquery
22 Recursive alge‐languages become Turin
bra with relational (query) algebra [Hutchison et al., 2017], and so forth. operations on binary arrays.
Knowledge graphs are often associated with a semantic schema or ontology that defines the seman‐
tics of domain terms, giving rise to entailments (per Chapter 4). Applying analytics with or without
such entailments – e.g., before or after materialisation – may yield radically different results. For ex‐
ample, observe that an edge Santa Lucía –hosts ➛ EID15 is semantically equivalent to an edge EID15 –
venue ➛ Santa Lucía once the inverse axiom hosts –inv. of ➛ venue is invoked; however, these edges are
far from equivalent from the perspective of analytical techniques that consider edge direction, for
which including one type of edge, or the other, or both, may have a major bearing on the final re‐
sults. To the best of our knowledge, the combination of analytics and entailment has not been well-
explored, leaving open interesting research questions. Along these lines, it may be of interest to ex‐
plore semantically-invariant analytics that yield the same results over semantically-equivalent graphs
(i.e., graphs that entail one another), thus analysing the semantic content of the knowledge graph
rather than simply the topological features of the data graph; for example, semantically-invariant an‐
alytics would yield the same results over a graph containing the inverse axiom hosts –inv. of ➛ venue
and a number of hosts edges, the same graph but where every hosts edge is replaced by an inverse
venue edge, and the union of both graphs.
Methods for machine learning have gained significant attention in recent years. In the context of
knowledge graphs, machine learning can either be used for directly refining a knowledge
graph [Paulheim, 2017] (discussed further in Chapter 8); or for downstream tasks using the knowl‐
edge graph, such as recommendation [Zhang et al., 2016], information extraction [Vashishth et al.,
2018], question answering [Huang et al., 2019], query relaxation [Wang et al., 2018], query approx‐
imation [Hamilton et al., 2018], etc. (discussed further in Chapter 10). However, many traditional
machine learning techniques assume dense numeric input representations in the form of vectors,
which is quite distinct from how graphs are usually expressed. So how can graphs – or nodes, edges,
etc., thereof – be encoded as numeric vectors? Give us feedback!
https://kgbook.org 55/150
2/15/23, 5:13 PM Knowledge Graphs
A first attempt to represent a graph using vectors would be to use a one-hot encoding, generating a
vector for each node of length ∣L∣ ⋅ ∣V ∣ – with ∣V ∣ the number of nodes in the input graph and ∣L∣ the
number of edge labels – placing a one at the corresponding index to indicate the existence of the re‐
spective edge in the graph, or zero otherwise. Such a representation will, however, typically result in
large and sparse vectors, which will be detrimental for most machine learning models.
The main goal of knowledge graph embedding techniques is to create a dense representation of the
graph (i.e., embed the graph) in a continuous, low-dimensional vector space that can then be used for
machine learning tasks. The dimensionality d of the embedding is fixed and usually low (often, e.g.,
50 ≥ d ≥ 1000). Typically the graph embedding is composed of an entity embedding for each node: a
vector with d dimensions that we denote by e; and a relation embedding for each edge label: (typical‐
ly) a vector with d dimensions that we denote by r. The overall goal of these vectors is to abstract
and preserve latent structures in the graph. There are many ways in which this notion of an embed‐
ding can be instantiated. Most commonly, given an edge s –p ➛ o , a specific embedding approach
defines a scoring function that accepts es (the entity embedding of node s ), rp (the entity embedding
of edge label p) and eo (the entity embedding of node o ) and computes the plausibility of the edge,
Table of Contents
which estimates how likely it is to be true. Given a data graph, the goal is then to compute the em‐
beddings of dimension d that maximise the plausibility of positive edges (typically edges in the
graph) and minimise the plausibility of negative examples (typically edges in the graph with a node
or edge label changed such that they are no longer in the graph) according to the given scoring func‐
tion. The resulting embeddings can then be seen as models learnt through self-supervision that en‐
code (latent) features of the graph, mapping input edges to output plausibility scores.
Embeddings can then be used for a number of low-level tasks involving the nodes and edge-labels of
the graph from which they were computed. First, we can use the plausibility scoring function to as‐
sign a confidence to edges that may, for example, have been extracted from an external source (dis‐
cussed later in Chapter 6). Second, the plausibility scoring function can be used to complete edges
with missing nodes/edge labels for the purposes of link prediction (discussed later in Chapter 8); for
example, in Figure 5.2, we might ask which nodes in the graph are likely to complete the edge
Grey Glacier –bus ➛ ? , where – aside from Punta Arenas , which is already given – we might intuitively
expect Torres del Paine to be a plausible candidate. Third, embedding models will typically assign sim‐
ilar vectors to similar nodes and similar edge-labels, and thus they can be used as the basis of simi‐
larity measures, which may be useful for finding duplicate nodes that refer to the same entity, or for
the purposes of providing recommendations (discussed later in Chapter 10).
A wide range of knowledge graph embedding techniques have been proposed [Wang et al., 2017].
Our goal here is to provide a high-level introduction to some of the most popular techniques pro‐
posed thus far. We first discuss tensor-based approaches that include three different sub-approaches
using linear/tensor algebra to compute embeddings. We then discuss language models that leverage
existing word embedding techniques, proposing ways of generating graph-like analogues for their ex‐
pected (textual) inputs. Finally we discuss entailment-aware models that can take into account the se‐
mantics of the graph, when available.
We first discuss tensor-based models, which we sub-divide into three categories: translational models
that adopt a geometric perspective whereby relation embeddings translate subject entities to object
entities, tensor decomposition models that extract latent factors approximating the graph’s structure,
and neural models that use neural networks to train embeddings that provide accurate plausibility
scores.
Translational models
Translational models interpret edge labels as transformations from subject nodes (aka the source or
head) to object nodes (aka the target or tail); for example, in the edge San Pedro –bus ➛ Moon Valley , the
edge label bus is seen as transforming San Pedro to Moon Valley , and likewise for other bus edges. The
most elementary approach in this family is TransE [Bordes et al., 2013]. Over all positive edges s –pGive us feedback!
https://kgbook.org 56/150
2/15/23, 5:13 PM Knowledge Graphs
➛ o , TransE learns vectors es, rp, and eos aiming to make es + rp as close as possible to eo.
Conversely, if the edge is a negative example, TransE attempts to learn a representation that keeps
es + rp away from eo. To illustrate, Figure 5.5 provides a toy example of two-dimensional (d = 2)
entity and relation embeddings computed by TransE. We keep the orientation of the vectors similar
to the original graph for clarity. For any edge s –p ➛ o in the original graph, adding the vectors
es + rp should approximate eo. In this toy example, the vectors correspond precisely where, for in‐
stance, adding the vectors for Licantén (eL.) and west of (rwo.) gives a vector corresponding to Curico
(eC.). We can use these embeddings to predict edges (amongst other tasks); for example, in order to
predict which node in the graph is most likely to be west of Antofagasta (A.), by computing
eA. + rwo. we find that the resulting vector (dotted in Figure 5.5c) is closest to eT., thus predicting
Toconao (T.) to be the most plausible such node.
Table of Contents
Figure 5.5: Toy example of two-dimensional relation and entity embeddings learnt by TransE; the entity
embeddings use abbreviations and include an example of vector addition to predict what is west of
Antofagasta
Aside from this toy example, TransE can be too simplistic; for example, in Figure 5.2, bus not only
transforms San Pedro to Moon Valley , but also to Arica , Calama , and so forth. TransE will, in this case,
aim to give similar vectors to all such target locations, which may not be feasible given other edges.
TransE will also tend to assign cyclical relations a zero vector, as the directional components will
tend to cancel each other out. To resolve such issues, many variants of TransE have been investigat‐
ed. Amongst these, for example, TransH [Wang et al., 2014] represents different relations using dis‐
tinct hyperplanes, where for the edge s –p ➛ o , s is first projected onto the hyperplane of p before
the translation to o is learnt (uninfluenced by edges with other labels for s and for o ). TransR [Lin
et al., 2015] generalises this approach by projecting s and o into a vector space specific to p, which
involves multiplying the entity embeddings for s and o by a projection matrix specific to p.
TransD [Ji et al., 2015] simplifies TransR by associating entities and relations with a second vector,
where these secondary vectors are used to project the entity into a relation-specific vector space.
Recently, RotatE [Sun et al., 2019] proposes translational embeddings in complex space, which al‐
lows to capture more characteristics of relations, such as direction, symmetry, inversion, antisymme‐
try, and composition. Embeddings have also been proposed in non-Euclidean; for example,
MuRP [Balazevic et al., 2019a] uses relation embeddings that transform entity embeddings in the
hyperbolic space of the Poincaré ball mode, whose curvature provides more “space” to separate enti‐
ties with respect to the dimensionality. For discussion of other translational models, we refer to sur‐
veys by Cai et al. [2018], Wang et al. [2017].
A second approach to derive graph embeddings is to apply methods based on tensor decomposition. A
tensor is a multidimensional numeric field that generalises scalars (0-order tensors), vectors (1-order
tensors) and matrices (2-order tensors) towards arbitrary dimension/order. Tensors have become a
widely used abstraction for machine learning [Rabanser et al., 2017]. Tensor decomposition involves
decomposing a tensor into more “elemental” tensors (e.g., of lower order) from which the original
tensor can be recomposed (or approximated) by a fixed sequence of basic operations over the output
tensors. These elemental tensors can be viewed as capturing latent factors underlying the information
Give us feedback!
https://kgbook.org 57/150
2/15/23, 5:13 PM Knowledge Graphs
contained in the original tensor. There are many approaches to tensor decomposition, where we will
now briefly introduce the main ideas behind rank decompositions [Rabanser et al., 2017].
Leaving aside graphs momentarily, consider an (a, b)-matrix (i.e., a 2-order tensor) C, where a is the
number of cities in Chile, b is the number of months in a year, and each element (C)ij denotes the
average temperature of the ith city in the j th month. Noting that Chile is a long, thin country – rang‐
ing from subpolar climates in the south, to a desert climate in the north – we may find a decomposi‐
tion of C into two vectors representing latent factors – specifically x (with a elements) giving lower
values for cities with lower latitude, and y (with b elements), giving lower values for months with
lower temperatures – such that computing the outer product23 of the two vectors approximates C
note 23 The outer product of two (column) vector
reasonably well: x ⊗ y ≈ C. In the (unlikely) case that there exist vectors x and y such that C is pre‐
x ⊗ y, is defined as xyT , yielding an (a, b)-
cisely the outer product of two vectors (x ⊗ y = C) we call C a rank-1 matrix; we can then precisely
Analogously, the outer product of k vectors is a
encode C using a + b values rather than a × b values. Most times, however, to get precisely C, we
need to sum multiple rank-1 matrices, where the rank r of C is the minimum number of rank-1 ma‐
trices that need to be summed to derive precisely C, such that x1 ⊗ y1 + … xr ⊗ yr = C. In the tem‐
perature example, x2 ⊗ y2 might correspond to a correction for altitude, x3 ⊗ y3 for higher tempera‐
Table of Contents
ture variance further south, etc. A (low) rank decomposition of a matrix then sets a limit d on the
rank and computes the vectors (x1 , y1 , … , xd , yd ) such that x1 ⊗ y1 + … + xd ⊗ yd gives the best d-
rank approximation of C. Noting that to generate n-order tensors we need to compute the outer
product of n vectors, we can generalise this idea towards low-rank decomposition of tensors; this
method is called Canonical Polyadic (CP) decomposition [Hitchcock, 1927]. For example, a 3-order
tensor C containing monthly temperatures for Chilean cities at four different times of day could be ap‐
proximated with x1 ⊗ y1 ⊗ z1 + … xd ⊗ yd ⊗ zd (e.g., x1 might be a latitude factor, y1 a monthly vari‐
ation factor, and z1 a daily variation factor, and so on). Various algorithms exist to compute (approx‐
imate) CP decompositions, including Alternating Least Squares, Jennrich’s Algorithm, and the Tensor
Power method [Rabanser et al., 2017].
Returning to graphs, similar principles can be used to decompose a graph into vectors, thus yielding
embeddings. In particular, a graph can be encoded as a one-hot 3-order tensor G with ∣V ∣ × ∣L∣ × ∣V ∣
elements, where the element (G)ijk is set to one if the ith node links to the k th node with an edge
having the j th label, or zero otherwise. As previously mentioned, such a tensor will typically be very
large and sparse, where rank decompositions are thus applicable. A CP decomposition [Hitchcock,
1927] would compute a sequence of vectors (x1 , y1 , z1 , … , xd , yd , zd ) such that x1 ⊗ y1 ⊗ z1 + … +
xd ⊗ yd ⊗ zd ≈ G . We illustrate this scheme in Figure 5.6. Letting X, Y, Z denote the matrices formed
by [x1 ⋯ xd ], [y1 ⋯ yd ], [z1 ⋯ zd ], respectively, with each vector forming a column of the corre‐
sponding matrix, we could then extract the ith row of Y as an embedding for the ith relation, and the
j th rows of X and Z as two embeddings for the j th entity. However, knowledge graph embeddings
typically aim to assign one vector to each entity.
Figure 5.6: Abstract illustration of a CP d-rank decomposition of a tensor representing the graph of
Figure 5.5a
DistMult [Yang et al., 2015] is a seminal method for computing knowledge graph embeddings based
on rank decompositions, where each entity and relation is associated with a vector of dimension d,
such that for an edge s –p ➛ o , a plausibility scoring function ∑di=1 (es)i (rp)i (eo)i is defined, where
(es)i , (rp)i and (eo)i denote the ith elements of vectors es, rp, eo, respectively. The goal, then, is to
learn vectors for each node and edge label that maximise the plausibility of positive edges and min‐
Give us feedback!
https://kgbook.org 58/150
2/15/23, 5:13 PM Knowledge Graphs
imise the plausibility of negative edges. This approach equates to a CP decomposition of the graph
tensor G , but where entities have one vector that is used twice: x1 ⊗ y1 ⊗ x1 + … + xd ⊗ yd ⊗ xd ≈ G .
A weakness of this approach is that per the scoring function, the plausibility of s –p ➛ o will always
be equal to that of o –p ➛ s ; in other words, DistMult does not consider edge direction.
Rather than use a vector as a relation embedding, RESCAL [Nickel and Tresp, 2013] uses a matrix,
which allows for combining values from es and eo across all dimensions, and thus can capture (e.g.)
edge direction. However, RESCAL incurs a higher cost in terms of space and time than DistMult.
HolE [Nickel et al., 2016b] uses vectors for relation and entity embeddings, but proposes to use the
circular correlation operator – which takes sums along the diagonals of the outer product of two vec‐
tors – to combine them. This operator is not commutative, and can thus consider edge direction.
ComplEx [Trouillon et al., 2016], on the other hand, uses a complex vector (i.e., a vector containing
complex numbers) as a relational embedding, which similarly allows for breaking the aforemen‐
tioned symmetry of DistMult’s scoring function while keeping the number of parameters low.
SimplE [Kazemi and Poole, 2018] rather proposes to compute a standard CP decomposition comput‐
ing two initial vectors for entities from X and Z and then averaging terms across X, Y, Z to com‐
Table of Contents
pute the final plausibility scores. TuckER [Balazevic et al., 2019b] employs a different type of de‐
composition – called a Tucker Decomposition [Tucker, 1964], which computes a smaller “core” ten‐
sor T and a sequence of three matrices A, B and C, such that G ≈ T ⊗ A ⊗ B ⊗ C – where entity
embeddings are taken from A and C, while relation embeddings are taken from B. Of these ap‐
proaches, TuckER [Balazevic et al., 2019b] currently provides state-of-the-art results on standard
benchmarks.
Neural models
A limitation of the aforementioned approaches is that they assume either linear (preserving addition
and scalar multiplication) or bilinear (e.g., matrix multiplication) operations over embeddings to
compute plausibility scores. Other approaches rather use neural networks to learn embeddings with
non-linear scoring functions for plausibility.
One of the earliest proposals of a neural model was Semantic Matching Energy (SME) [Glorot et al.,
2013], which learns parameters (aka weights: w, w′ ) for two functions – fw (es, rp) and gw′ (eo, rp) –
such that the dot product of the result of both functions – fw (es, rp) ⋅ gw′ (eo, rp) – gives the plausibili‐
ty score. Both linear and bilinear variants of fw and gw′ are proposed. Another early proposal was
Neural Tensor Networks (NTN) [Socher et al., 2013], which proposes to maintain a tensor W of in‐
ternal weights, such that the plausibility score is computed by a complex function that combines the
outer product es⊗W ⊗ eo with a standard neural layer over es and eo, which in turn is combined
with rp, to produce a plausibility score. The tensor W results in a high number of parameters, limit‐
ing scalability [Wang et al., 2017]. Multi-Layer Perceptron (MLP) [Dong et al., 2014] is a simpler
model, where es, rp and eo are concatenated and fed into a hidden layer to compute plausibility
scores.
A number of more recent approaches have proposed using convolutional kernels in their models.
ConvE [Dettmers et al., 2018] proposes to generate a matrix from es and rp by “wrapping” each vec‐
tor over several rows and concatenating both matrices. The concatenated matrix serves as the input
for a set of (2D) convolutional layers, which returns a feature map tensor. The feature map tensor is
vectorised and projected into d dimensions using a parameterised linear transformation. The plausi‐
bility score is then computed based on the dot product of this vector and eo. A disadvantage of
ConvE is that by wrapping vectors into matrices, it imposes an artificial two-dimensional structure
on the embeddings. HypER [Balazevic et al., 2019c] is a similar model using convolutions, but
avoids the need to wrap vectors into matrices. Instead, a fully connected layer (called the “hypernet‐
work”) is applied to rp and used to generate a matrix of relation-specific convolutional filters. These
filters are applied directly to es to give a feature map, which is vectorised. The same process is then
applied as in ConvE: the resulting vector is projected into d dimensions, and a dot product applied
with eo to produce the plausibility score. The resulting model is shown to outperform ConvE on stan‐
dard benchmarks [Balazevic et al., 2019c]. Give us feedback!
https://kgbook.org 59/150
2/15/23, 5:13 PM Knowledge Graphs
The presented approaches strike different balances in terms of expressivity and the number of para‐
meters than need to be trained. While more expressive models, such as NTN, may better fit more
complex plausibility functions over lower dimensional embeddings by using more hidden parame‐
ters, simpler models, such as that proposed by Dong et al. [Dong et al., 2014], and convolutional net‐
works [Dettmers et al., 2018, Balazevic et al., 2019c] that enable parameter sharing by applying the
same (typically small) kernels over different regions of a matrix, require handling fewer parameters
overall and are more scalable.
We now formally define and survey the aforementioned tensor-based approaches. For simplicity, we
will consider directed edge-labelled graphs.
For any positive integer a, a vector of dimension a is a family of real numbers indexed by inte‐
gers in {1, … , a}. For a and b positive integers, an (a, b)-matrix is a family of real numbers in‐
dexed by pairs of integers in {1, … , a} × {1, … , b}. A tensor is a family of real numbers indexed
by a finite sequence of integers such that there exist positive numbers a1 , … , an such that the in‐
dices are all the tuples of numbers in {1, … , a1 } × … × {1, … , an }. The number n is called the
order of the tensor, the subindices i ∈ {1, … , n} indicate the mode of a tensor, and each ai de‐
fines the dimension of the i th mode. A 1-order tensor is a vector and a 2-order tensor is a matrix.
We denote the set of all tensors as T.
For specific dimensions a1 , … , an of modes, a tensor is an element of (⋯ (Ra1 )… )an but we write
R a1 ,…,an
to simplify the notation. We use lower-case bold font to denote vectors (x ∈ Ra ), upper-
case bold font to denote matrices (X ∈ Ra,b ) and calligraphic font to denote tensors
(X ∈ Ra1 ,…,an ).
In the most typical case, ε and ρ map nodes and edge-labels, respectively, to vectors of fixed di‐
mension. In some cases, however, they may map to matrices. Given this abstract notion of a
knowledge graph embedding, we can then define a plausibility scoring function.
Edges with higher scores are considered more plausible. Given a graph G = (V , E, L), we as‐
sume a set of positive edges E + and a set of negative edges E − . Positive edges are often simply
the edges in the graph: E + := E . Negative edges use the vocabulary of G (i.e., E − ⊆ V × L × V )
and are typically defined by taking edges (s, p, o) from E and changing one term of each edge –
often one of the nodes – such that the edge is no longer in E . Given sets of positive and negative
edges, and a plausibility scoring function, the objective is then to find the embedding that max‐
imises the plausibility of edges in E + while minimising the plausibility of edges in E − . Specific
knowledge graph embeddings then instantiate the type of embedding considered and the plausi‐
bility scoring function in various ways.
In Table 5.1, we define the plausibility scoring function and types of embeddings used by differ‐
ent knowledge graph embeddings. To simplify the definitions, we use ex to denote ε(x) when it
Give us feedback!
https://kgbook.org 60/150
2/15/23, 5:13 PM Knowledge Graphs
is a vector, ry to denote ρ(y) when it is a vector, and Ry to denote ρ(y) when it is a matrix.
Some models involve learnt parameters (aka weights) for computing plausibility. We denote
these as v, V, V , w, W W (for vectors, matrices or tensors). We use de and dr to denote the di‐
mensionality chosen for entity embeddings and relation embeddings, respectively. Often it is as‐
sumed that de = dr , in which case we will write d. Weights may have their own dimensionality,
which we denote w. The embeddings in Table 5.1 use a variety of operators on vectors, matrices
and tensors, which will be defined later.
The embeddings defined in Table 5.1 vary in complexity, where a trade-off exists between the
number of parameters used, and the expressiveness of the model in terms of its capability to
capture latent features of the graph. To increase expressivity, many of the models in Table 5.1
use additional parameters beyond the embeddings themselves. A possible formal guarantee of
such models is full expressiveness, which, given any disjoint sets of positive edges E + and nega‐
tive edges E − , asserts that the model can always correctly partition those edges. On the one
hand, for example, DistMult [Yang et al., 2015] cannot distinguish an edge s –p ➛ o from its in‐
verse o –p ➛ s , so by adding an inverse of an edge in E + to E − , we can show that it is not fully
Table of Contents
expressive. On the other hand, models such as ComplEx [Trouillon et al., 2016], SimplE [Kazemi
and Poole, 2018], and TuckER [Balazevic et al., 2019b] have been proven to be fully expressive
given sufficient dimensionality; for example, TuckER [Balazevic et al., 2019b] with dimensions
dr = ∣L∣ and de = ∣V ∣ trivially satisfies full expressivity since its core tensor W then has sufficient
capacity to store the full one-hot encoding of any graph. This formal property is useful to show
that the model does not have built-in limitations for numerically representing a graph, though of
course in practice the dimensions needed to reach full expressivity are often
impractical/undesirable.
We use (x)i , (X)ij , and (X )i1 …in to denote elements of vectors, matrices, and tensors, re‐
spectively. If a vector x ∈ Ra is used in a context that requires a matrix, the vector is inter‐
preted as an (a, 1)-matrix (i.e., a column vector) and can be turned into a row vector (i.e.,
a (1, a)-matrix) using the transpose operation xT . We use xD ∈ Ra,a to denote the diagonal
matrix with the values of the vector x ∈ Ra on its diagonal. We denote the identity matrix
by I such that if j = k , then (I)jk = 1; otherwise (I)jk = 0.
⎡ X1 ⎤
columns. Given a vector x ∈ Rab , we denote by x[a,b] ∈ Ra,b the “reshaping” of x into an
(a, b)-matrix such that (x[a,b] )ij = (x)(i+a(j−1)) . Conversely, given a matrix X ∈ Ra,b , we de‐
note by vec(X) ∈ Rab the vectorisation of X such that vec(X)k = (X)ij where i = ((k −
Given a tensor X ∈ Ra,b,c , we denote by X [i:⋅:⋅] ∈ Rb,c , the ith slice of tensor X along the first
mode; for example, given X ∈ R5,2,3 , then X [4:⋅:⋅] returns the (2, 3)-matrix consisting of the
(X )411 (X )412 (X )413
elements [ ]. Analogously, we use X [⋅:i:⋅] ∈ Ra,c and X [⋅:⋅:i] ∈ Rb,c to in‐
dicate the ith slice along the second and third modes of X , respectively.
We denote by ψ(X ) the element-wise application of a function ψ to the tensor X , such that
(ψ(X ))in1 …in = ψ(Xi1 …in ). Common choices for ψ include a sigmoid function (e.g., the lo‐
We now define the operators used in Table 5.1, where the first and most elemental operation we
consider is that of matrix multiplication.
R c,b1 ,…,bn
Give us feedback!
https://kgbook.org 61/150
2/15/23, 5:13 PM Knowledge Graphs
For convenience, we may implicitly add or remove modes with dimension 1 for the purposes of
matrix multiplication and other operators; for example, given two vectors x ∈ Ra and y ∈ Ra , we
denote by xT y (aka the dot or inner product) the multiplication of matrix xT ∈ R1,a with y ∈
Ra,1 such that xT y ∈ R1,1 (i.e., a scalar in R); conversely, xyT ∈ Ra,a (the outer product).
∣(x)i ∣ denotes the absolute value of the ith element of x. For p, q ∈ R, the Lp,q -norm of a matrix
The L1 norm (i.e., ∥x∥1 ) is thus simply the sum of the absolute values of x, while the L2 norm
(i.e., ∥x∥2 ) is the (Euclidean) length of the vector. The Frobenius norm of the matrix X then
equates to ∥X∥2,2 = (∑bj=1 (∑ai=1 ∣(X)ij ∣2 )) ; i.e., the square root of the sum of the squares of all
2
elements.
Table of Contents
Another type of product used by embedding techniques is the Hadamard product, which multi‐
plies tensors of the same dimension and computes their product in an element-wise manner.
tensor in R a1 ,…,an
, with each element computed as (X ⊙ Y)i1 …in := (X )i1 …in (Y)i1 …in .
Other embedding techniques – namely RotatE [Sun et al., 2019] and ComplEx [Trouillon et al.,
2016] – uses complex space based on complex numbers. With a slight abuse of notation, the defi‐
nitions of vectors, matrices and tensors can be modified by replacing the set of real numbers R
by the set of complex numbers C, giving rise to complex vectors, complex matrices, and complex
tensors. In this case, we denote by Re(⋅) the real part of a complex number. Given a complex
vector x ∈ CI , we denote by x its complex conjugate (swapping the sign of the imaginary part of
each element). Complex analogues of the aforementioned operators can then be defined by re‐
placing the multiplication and addition of real numbers with the analogous operators for com‐
plex numbers, where RotateE [Sun et al., 2019] uses the complex Hadamard product, and
ComplEx [Trouillon et al., 2016] uses complex matrix multiplication.
One embedding technique – MuRP [Balazevic et al., 2019a] – uses hyperbolic space, specifically
based on the Poincaré ball. As this is the only embedding we cover that uses this space, and the
formalisms are lengthy (covering the Poincaré ball, Möbius addition, Möbius matrix–vector mul‐
tiplication, logarithmic maps, exponential maps, etc.), we rather refer the reader to the paper for
further details [Balazevic et al., 2019a].
As discussed in Section 5.2, tensor decompositions are used for many embeddings, and at the
heart of such decompositions is the tensor product, which is often used to reconstruct (an ap‐
proximation of) the original tensor.
sor in Ra1 ,…,am ,b1 ,…,bn , with each element computed as (X ⊗ Y)i1 …im j1 …jn := (X 24
note
)i1 …24
im (Y) j1 …jn . note that “⊗” is used here
Please
Definition 3.10.
Example 5.3 — Assume that X ∈ R2,3 and Y ∈ R3,4,5 . Then X ⊗ Y will be a tensor in R2,3,3,4,5 .
Element (X ⊗ Y)12345 will be the product of (X )12 and (Y)345 .
An n-mode product is used by other embeddings to transform elements along a given mode of a
tensor by computing a product with a given matrix along that particular mode of the tensor.
Give us feedback!
https://kgbook.org 62/150
2/15/23, 5:13 PM Knowledge Graphs
such that (X ⊗n
Example 5.4 — Let us assume that X ∈ R2,3,4 and Y ∈ R5,3 . The result of X ⊗2 Y will be a tensor in
R 2,5,4
, where, for example, (X ⊗2 Y)142 will be given as (X )112 (Y)41 + (X )122 (Y)42 + (X )132 (Y)43 .
Observe that if y ∈ Ran – i.e., if y is a (column) vector – then the n-mode tensor product X ⊗n yT
“flattens” the nth mode of X to one dimension, effectively reducing the order of X by one.
One embedding technique – HolE [Nickel et al., 2016b] – uses the circular correlation operator
x ⋆ y, where each element is the sum of elements along a diagonal of the outer product x ⊗ y
that “wraps” if not the primary diagonal.
Example 5.5 — Assuming a = 5, then (x ⋆ y)1 = (x)1 (y)1 + (x)2 (y)2 + (x)3 (y)3 + (x)4 (y)4 +
(x)5 (y)5 , or a case that wraps: (x ⋆ y)4 = (x)1 (y)4 + (x)2 (y)5 + (x)3 (y)1 + (x)4 (y)2 + (x)5 (y)3 .
Finally, a couple of neural models that we include – namely ConvE [Dettmers et al., 2018] and
HypER [Balazevic et al., 2019c] – are based on convolutional architectures using the convolu‐
tion operator.
tional
k − a) < 1, (j + l − b) < 1, (i + k − a) > e or (j + l − b) > f (i.e., where (Y) neural networks.
(i+k−a)(j+l−b) lies
Strictly speaking, the o
where traditional convolution requires the ma
lapping elements to generate an element of the result. Elements of X extending beyond Y are
ignored (equivalently we can consider Y to be “zero-padded” outside its borders).
Example 5.6 — Given X ∈ R3,3 and Y ∈ R4,5 , then X ∗ Y ∈ R6,7 , where, for example, (X ∗ Y)11 =
(X)33 (Y)11 (with the bottom right corner of X overlapping the top left corner of Y), while (X ∗
Y)34 = (X)11 (Y)12 +(X)12 (Y)13 +(X)13 (Y)14 +(X)21 (Y)22 +(X)22 (Y)23 +(X)23 (Y)24 +(X)31 (Y)32 +
Table 5.1: Details for selected knowledge graph embeddings, including the plausibility scoring function
ϕ(ε(s), ρ(p), ε(o)) for edge s –p ➛ o , and other conditions
Model ϕ(ε(s), ρ(p), ε(o)) Conditions (for all x ∈ V , y ∈ L)
TransE −∥es + rp − eo ∥q
ex ∈ Rd , ry ∈ Rd , q ∈ {1, 2}, ∥ex ∥2 = 1
ex ∈ R d , r y ∈ R d , w y ∈ R d ,
TransH −∥(es − (eT T 2
∥ry ∥2
ex ∈ R , r y ∈ R , W y ∈ R d r , d e , de dr
ex ∈ R d e , r y ∈ R d r , w x ∈ R d e , w y ∈ R d r ,
RotatE −∥es ⊙ rp − eo ∥2
ex ∈ Cd , ry ∈ Cd , ∥ry ∥2 = 1
Give us feedback!
https://kgbook.org 63/150
2/15/23, 5:13 PM Knowledge Graphs
DistMult eT D
s r p eo ex ∈ Rd , ry ∈ Rd , ∥ex ∥2 = 1, ∥ry ∥2 ≤ 1
HolE rT
p ( es ⋆ eo )
ex ∈ Rd , ry ∈ Rd , ∥ex ∥2 ≤ 1, ∥ry ∥2 ≤ 1
ComplEx Re(eT D
s r p eo ) ex ∈ Cd , ry ∈ Cd , ∥ex ∥2 ≤ 1, ∥ry ∥2 ≤ 1
eT D T D
s r p wo + e o wp ws
ex ∈ R d , r y ∈ R d , w x ∈ R d , w y ∈ R d ,
SimplE
2
∥ex ∥2 ≤ 1, ∥wx ∥2 ≤ 1, ∥ry ∥2 ≤ 1, ∥wy ∥2 ≤ 1
TuckER W ⊗ 1 eT
s ⊗2 r T
p ⊗ 3 eT
o ex ∈ R d e , r y ∈ R d r , W ∈ R d e , d r , d e
ex ∈ Rd , ry ∈ Rd , v ∈ Rw , w ∈ Rw , ∥ex ∥2 = 1,
SME L. (Ves + V′ rp + v)T (Weo + W′ rp + w)
ex ∈ Rd , ry ∈ Rd , v ∈ Rw , w ∈ Rw , ∥ex ∥2 = 1,
SME Bi. ((V ⊗3 rT T T
p )es + v) ((W ⊗3 rp )eo + w)
V ∈ Rw,d,d , W ∈ Rw,d,d
ex ∈ Rd , ry ∈ Rd , w ∈ Rw , W ∈ Rw,2d ,
es
NTN p ψ (es Weo + W [e ] + w)
rT W ∈ Rd,w,d , ∥ex ∥2 ≤ 1, ∥ry ∥2 ≤ 1,
T
o
⎡ es ⎤ ⎞ ⎛
ex ∈ Rd , ry ∈ Rd , v ∈ Rw , w ∈ Rw , W ∈ Rw,3d
MLP v T ψ W rp + w
⎝ ⎣ eo ⎦ ⎠
∥ex ∥2 ≤ 1 ∥ry ∥2 ≤ 1
⎛ ⎞
T
e[a,b] ex ∈ Rd , ry ∈ Rd , d = ab,
ψ vec (ψ (W ∗ [ [a,b] ])) W eo
Table of Contents
ConvE
s
⎝ ⎠
ex ∈ R , r y ∈ R , W ∈ R
de dr w2 (w1 +de −1),de
,
ψ (vec (rT
p W ∗ es ) W ) eo
T
HypER
W ∈ Rd r , w 1 , w 2
Embedding techniques were first explored as a way to represent natural language within machine
learning frameworks, with word2vec [Mikolov et al., 2013] and GloVe [Pennington et al., 2014] be‐
ing two seminal approaches. Both approaches compute embeddings for words based on large corpora
of text such that words used in similar contexts (e.g., “frog”, “toad”) have similar vectors.
Word2vec uses neural networks trained either to predict the current word from surrounding words
(continuous bag of words), or to predict the surrounding words given the current word (continuous
skip-gram). GloVe rather applies a regression model over a matrix of co-occurrence probabilities of
word pairs. Embeddings generated by both approaches have become widely used in natural language
processing tasks.
Another approach for graph embeddings is thus to leverage proven approaches for language embed‐
dings. However, while a graph consists of an unordered set of sequences of three terms (i.e., a set of
edges), text in natural language consists of arbitrary-length sequences of terms (i.e., sentences of
words). RDF2Vec [Ristoski and Paulheim, 2016] thus performs (biased [Cochez et al., 2017a]) ran‐
dom walks on the graph and records the paths (the sequence of nodes and edge labels traversed) as
“sentences”, which are then fed as input into the word2vec [Mikolov et al., 2013] model. An exam‐
ple of such a path extracted from Figure 5.2 might be, for example, San Pedro –bus ➛ Calama –flight ➛
Iquique –flight ➛ Santiago , where the paper experiments with 500 paths of length 8 per entity. RDF2Vec
also proposes a second mode where sequences are generated for nodes from canonically-labelled sub-
trees of which they are a root node, where sub-trees of depth 1 and 2 are used for experiments.
KGloVe [Cochez et al., 2017b] is rather based on GloVe. Given that the original GloVe
model [Pennington et al., 2014] considers words that co-occur frequently in windows of text to be
more related, KGloVe uses personalised PageRank27 to determine the most related
note 27 nodes to a speaking,
Intuitively given personalised PageRa
node, which are fed into the GloVe model. mines the probability of a random walk being a
of steps. A higher number of steps converges
global node centrality in the graph, while a low
5.2.3 Entailment-aware models ness to the starting node.
The embeddings thus far consider the data graph alone. But what if an ontology or set of rules is pro‐
vided? Such deductive knowledge could be used to improve the embeddings. One approach is to use
constraint rules to refine the predictions made by embeddings; for example, Wang et al. [2015] use
functional and inverse-functional definitions as constraints (under UNA) such that, for example, if we
define that an event can have at most one value for venue, this is used to lower the plausibility of
edges that would assign multiple venues to an event.
Give us feedback!
https://kgbook.org 64/150
2/15/23, 5:13 PM Knowledge Graphs
More recent approaches rather propose joint embeddings that consider both the data graph and rules
when computing embeddings. KALE [Guo et al., 2016] computes entity and relation embeddings us‐
ing a translational model (specifically TransE) that is adapted to further consider rules using t-norm
fuzzy logics. With reference to Figure 5.2, consider a simple rule ?x –bus ➛ ?y ⇒ ?x –connects to ➛ ?y .
We can use embeddings to assign plausibility scores to new edges, such as e1 : Piedras Rojas –bus ➛
Moon Valley . We can further apply the previous rule to generate a new edge e2 : Piedras Rojas –
connects to ➛ Moon Valley from the predicted edge e1 . But what plausibility should we assign to this sec‐
ond edge? Letting p1 and p2 be the current plausibility scores of e1 and e2 (initialised using the stan‐
dard embedding), then t-norm fuzzy logics suggests that the plausibility be updated as p1 p2 − p1 + 1.
Embeddings are then trained to jointly assign larger plausibility scores to positive examples versus
negative examples of both edges and ground rules. An example of a positive ground rule based on
Figure 5.2 would be Arica –bus ➛ San Pedro ⇒ Arica –connects to ➛ San Pedro . Negative ground rules ran‐
domly replace the relation in the head of the rule; for example, Arica –bus ➛ San Pedro
⇒ Arica –flight
➛ San Pedro . Guo et al. [2018] later propose RUGE, which uses a joint model over ground rules (pos‐
sibly soft rules with confidence scores) and plausibility scores to align both forms of scoring for un‐
seen edges.
Table of Contents
Generating ground rules can be costly. An alternative approach, called FSL [Demeester et al., 2016],
observes that in the case of a simple rule, such as ?x –bus ➛ ?y ⇒ ?x –connects to ➛ ?y , the relation
embedding bus should always return a lower plausibility than connects to. Thus, for all such rules,
FSL proposes to train relation embeddings while avoiding violations of such inequalities. While rela‐
tively straightforward, FSL only supports simple rules, while KALE also supports more complex rules.
These works exemplify how deductive and inductive forms of knowledge – in this case rules and em‐
beddings – can interplay and complement each other.
While embeddings aim to provide a dense numerical representation of graphs suitable for use within
existing machine learning models, another approach is to build custom machine learning models
adapted for graph-structured data. Most custom learning models for graphs are based on (artificial)
neural networks [Wu et al., 2019], exploiting a natural correspondence between both: a neural net‐
work already corresponds to a weighted, directed graph, where nodes serve as artificial neurons, and
edges serve as weighted connections (axons). However, the typical topology of a traditional neural
network – more specifically, a fully-connected feed-forward neural network – is quite homogeneous,
being defined in terms of sequential layers of nodes where each node in one layer is connected to all
nodes in the next layer. Conversely, the topology of a data graph is quite heterogeneous, being deter‐
mined by the relations between entities that its edges represent.
A graph neural network (GNN) [Scarselli et al., 2009] builds a neural network based on the topology
of the data graph; i.e., nodes are connected to their neighbours per the data graph. Typically a model
is then learnt to map input features for nodes to output features in a supervised manner; output fea‐
tures of the example nodes used for training may be manually labelled, or may be taken from the
knowledge graph. Unlike knowledge graph embeddings, GNNs support end-to-end supervised learn‐
ing for specific tasks: given a set of labelled examples, GNNs can be used to classify elements of the
graph or the graph itself. GNNs have been used to perform classification over graphs encoding com‐
pounds, objects in images, documents, etc.; as well as to predict traffic, build recommender systems,
verify software, etc. [Wu et al., 2019]. Given labelled examples, GNNs can even replace graph algo‐
rithms; for example, GNNs have been used to find central nodes in knowledge graphs in a supervised
manner [Scarselli et al., 2009, Park et al., 2019, Park et al., 2020].
We now discuss the ideas underlying two main flavours of GNN, specifically, recursive GNNs and non-
recursive GNNs.
Give us feedback!
https://kgbook.org 65/150
2/15/23, 5:13 PM Knowledge Graphs
Recursive graph neural networks (RecGNNs) are the seminal approach to graph neural
networks [Sperduti and Starita, 1997, Scarselli et al., 2009]. The approach is conceptually similar to
the systolic abstraction illustrated in Figure 5.3, where messages are passed between neighbours to‐
wards recursively computing some result. However, rather than define the functions used to decide
the messages to pass, we rather label the output of a training set of nodes and let the framework
learn the functions that generate the expected output, thereafter applying them to label other
examples.
In a seminal paper, Scarselli et al. [2009] proposed what they generically call a graph neural net‐
work (GNN), which takes as input a directed graph where nodes and edges are associated with fea‐
ture vectors that can capture node and edge labels, weights, etc. These feature vectors remain fixed
throughout the process. Each node in the graph is also associated with a state vector, which is recur‐
sively updated based on information from the node’s neighbours – i.e., the feature and state vectors
of the neighbouring nodes and the feature vectors of the edges extending to/from them – using a
parametric function, called the transition function. A second parametric function, called the output
function, is used to compute the final output for a node based on its own feature and state vector.
Table of Contents
These functions are applied recursively up to a fixpoint. Both parametric functions can be imple‐
mented using neural networks where, given a partial set of supervised nodes in the graph – i.e., nodes
labelled with their desired output – parameters for the transition and output functions can be learnt
that best approximate the supervised outputs. The result can thus be seen as a recursive neural net‐
work architecture.28 To ensure convergence up to a fixpoint, certain restrictions
note are applied,
28 Some namely
authors refer to such architectures
ing thatofthe
that the transition function be a contractor, meaning that upon each application theinternal state maintained for node
function,
over a sequence of transitions.
points in the numeric space are brought closer together (intuitively, in this case, the numeric space
“shrinks” upon each application, ensuring convergence to a unique fixpoint).
To illustrate, consider, for example, that we wish to find priority locations for creating new tourist
information offices. A good strategy would be to install them in hubs from which many tourists visit
popular destinations. Along these lines, in Figure 5.7 we illustrate the GNN architecture proposed by
Scarselli et al. [2009] for a sub-graph of Figure 5.2, where we highlight the neighbourhood of
Punta Arenas . In this graph, nodes are annotated with feature vectors (nx ) and hidden states at step t
(h(t)
x ), while edges are annotated with feature vectors (axy ). Feature vectors for nodes may, for exam‐
ple, one-hot encode the type of node (City, Attraction, etc.), directly encode statistics such as the
number of tourists visiting per year, etc. Feature vectors for edges may, for example, one-hot encode
the edge label (the type of transport), directly encode statistics such as the distance or number of
tickets sold per year, etc. Hidden states can be randomly initialised. The right-hand side of Figure 5.7
provides the GNN transition and output functions, where N(x) denotes the neighbouring nodes of x,
fw (⋅) denotes the transition function with parameters w, and gw′ (⋅) denotes the output function with
parameters w′ . An example is also provided for Punta Arenas (x = 1). These functions will be recur‐
sively applied until a fixpoint is reached. To train the network, we can label examples of places that
already have (or should have) tourist offices and places that do (or should) not have tourist offices.
These labels may be taken from the knowledge graph, or may be added manually. The GNN can then
learn parameters w and w′ that give the expected output for the labelled examples, which can subse‐
quently be used to label other nodes.
This GNN model is flexible and can be adapted in various ways [Scarselli et al., 2009]: we may de‐
fine neighbouring nodes differently, for example to include nodes for outgoing edges, or nodes one
or two hops away; we may allow pairs of nodes to be connected by multiple edges with different
vectors; we may consider transition and output functions with distinct parameters for each node; we
may add states and outputs for edges; we may change the sum to another aggregation function; etc.
(t) (t−1)
hx := ∑y∈N(x) fw (nx , ny , ayx , hy )
(t) (t)
ox := gw′ (hx , nx )
(t) (t−1)
h1 := fw (n1 , n3 , a31 , h3
)
(t−1)
+fw (n1 , n4 , a41 , h4
)
(t) (t)
o1 := gw′ (h1 , n1 )
… Give us feedback!
https://kgbook.org 66/150
2/15/23, 5:13 PM Knowledge Graphs
Figure 5.7: On the left a sub-graph of Figure 5.2 highlighting the neighbourhood of Punta Arenas,
where nodes are annotated with feature vectors (nx ) and hidden states at step t (h(t)
x ), and edges are anno‐
tated with feature vectors (axy ); on the right, the GNN transition and output functions proposed by
Scarselli et al. [2009] and an example for Punta Arenas (x = 1), where N(x) denotes the neighbouring
nodes of x, fw (⋅) denotes the transition function with parameters w and gw′ (⋅) denotes the output function
with parameters w′
We now define a recursive graph neural network. We assume that the GNN accepts a directed
vector-labelled graph as input (see Definition 5.1).
Out: R → R a c
The function Agg computes a new feature vector for a node, given its previous feature vector
and the feature vectors of the nodes and edges forming its neighbourhood; the function Out
transforms the final feature vector computed by Agg for a node to the output vector for that
node. We assume that a and b correspond to the dimensions of the input node and edge vectors,
respectively, while c denotes the dimension of the output vector for each node. Given a RecGNN
R = (Agg, Out), a directed vector-labelled graph G = (V , E, F , λ), and a node u ∈ V , we define
the output vector assigned to node u in G by R (written R(G, u)) as follows. First let n(0)
u := λ(u)
In a RecGNN, the same aggregation function (Agg) is applied recursively until a fixpoint is
reached, at which point an output function (Out}) creates the final output vector for each node.
While in practice RecGNNs will often consider a static feature vector and a dynamic state
vector [Scarselli et al., 2009], we can more concisely encode this as one vector, where part may
remain static throughout the aggregation process representing input features, and part may be
dynamically computed representing the state. In practice, Agg and Out are often based on para‐
metric combinations of vectors, with the parameters learnt based on a sample of output vectors
for labelled nodes.
Example 5.7 — The aggregation function for the GNN of Scarselli et al. [2009] is given as:
where fw (⋅) is a contraction function with parameters w. The output function is defined as:
where again gw′ (⋅) is a function with parameters w′ . Given a set of nodes labelled with their expected
There are notable similarities between graph parallel frameworks (GPFs; see Definition 5.2) and
RecGNNs. While we defined GPFs using separate Msg and Agg functions, this is not essential:
conceptually they could be defined in a similar way to RecGNN, with a single Agg function that
“pulls” information from its neighbours (we maintain Msg to more closely reflect how GPFs are
defined/implemented in practice). The key difference between GPFs and GNNs is that in the for‐
mer, the functions are defined by the user, while in the latter, the functions are generally learnt
from labelled examples. Another difference arises from the termination condition present in
Give us feedback!
https://kgbook.org 67/150
2/15/23, 5:13 PM Knowledge Graphs
GPFs, though often the GPF’s termination condition will – like in RecGNNs – reflect convergence
to a fixpoint.
GNNs can also be defined in a non-recursive manner, where a fixed number of layers are applied
over the input in order to generate the output. A benefit of this approach is that we do not need to
worry about convergence since the process is non-recursive. Also, each layer will often have inde‐
pendent parameters, representing different transformation steps. Naively, a downside is that adding
many layers could give rise to a high number of parameters. Addressing this problem, a popular ap‐
proach for non-recursive GNNs is to use convolutional neural networks.
Convolutional neural networks (CNNs) have gained a lot of attention, in particular, for machine
learning tasks involving images [Krizhevsky et al., 2017]. The core idea in the image setting is to
train and apply small kernels (aka filters) over localised regions of an image using a convolution op‐
Table of Contents
erator to extract features from that local region. When applied to all local regions, the convolution
outputs a feature map of the image. Since the kernels are small, and are applied multiple times to
different regions of the input, the number of parameters to train is reduced. Typically multiple ker‐
nels can thus be applied, forming multiple convolutional layers.
One may note that in GNNs and CNNs, operators are applied over local regions of the input data. In
the case of GNNs, the transition function is applied over a node and its neighbours in the graph. In
the case of CNNs, the convolution is applied on a pixel and its neighbours in the image. Following
this intuition, a number of convolutional graph neural networks (ConvGNNs) [Bruna et al., 2014, Kipf
and Welling, 2017, Wu et al., 2019] have been proposed, where the transition function is imple‐
mented by means of convolutions. A key consideration for ConvGNNs is how regions of a graph are
defined. Unlike the pixels of an image, nodes in a graph may have varying numbers of neighbours.
This creates a challenge: a benefit of CNNs is that the same kernel can be applied over all the regions
of an image, but this requires more careful consideration in the case of ConvGNNs since neighbour‐
hoods of different nodes can be diverse. Approaches to address these challenges involve working
with spectral (e.g. [Bruna et al., 2014, Kipf and Welling, 2017]) or spatial (e.g., [Monti et al., 2017])
representations of graphs that induce a more regular structure from the graph. An alternative is to
use an attention mechanism [Velickovic et al., 2018] to learn the nodes whose features are most im‐
portant to the current node.
Agg(k) Ra k .
a b
: Rak−1 × 2(R k−1 ×R )→N →
Each function Agg(k) (as before) computes a new feature vector for a node, given its previous
feature vector and the feature vectors of the nodes and edges forming its neighbourhood. We as‐
sume that a0 and b correspond to the dimensions of the input node and edge vectors, respective‐
ly, where each function Agg(k) for 2 ≤ k ≤ l accepts as input node vectors of the same dimension
as the output of the function Agg(k−1) . Given an NRecGNN N = ( Agg(1) , … , Agg(l) ), a directed
vector-labelled graph G = (V , E, F , λ), and a node u ∈ V , we define the output vector assigned
to node u in G by N (written N(G, u)) as follows. First let n(0)
u := λ(u). For all i ≥ 1, let:
Then N(G, u) := nu .
(l)
In an l-layer NRecGNN, a different aggregation function can be applied at each step (i.e., in each
layer), up to a fixed number of steps l. We do not consider a separate Out function as it can be
combined with the final aggregation function Agg(l) . When the aggregation functions use a con‐ Give us feedback!
https://kgbook.org 68/150
2/15/23, 5:13 PM Knowledge Graphs
volutional operator based on kernels learned from labelled examples, we call the result a convo‐
lutional graph neural network (ConvGNN). We refer to the survey by Wu et al. [2019] for discus‐
sion of ConvGNNs proposed in the literature.
We have considered GNNs that define the neighbourhood of a node based on its incoming edges.
These definitions can be adapted to also consider outgoing neighbours by either adding inverse
edges to the directed vector-labelled graph in pre-processing, or by adding outgoing neighbours
as arguments to the Agg(⋅) function. More generally, GNNs (and indeed GPFs) relying solely on
the neighbourhood of each node have limited expressivity in terms of their ability to distinguish
nodes and graphs [Xu et al., 2019]; for example, Barceló et al. [2020] show that such
NRecGNNs have a similar expressiveness for classifying nodes as the ALCQ Description Logic
discussed in Section 4.3.2. More expressive GNN variants have been proposed that allow the ag‐
gregation functions to access and update a globally shared vector [Barceló et al., 2020]. We re‐
fer to the papers by Xu et al. [2019] and Barceló et al. [2020] for further discussion.
Table of Contents
The supervised techniques discussed thus far – namely knowledge graph embeddings and graph
neural networks – learn numerical models over graphs. However, such models are often difficult to
explain or understand. For example, taking the graph of Figure 5.8, knowledge graph embeddings
might predict the edge SCL –flight ➛ ARI as being highly plausible, but they will not provide an inter‐
pretable model to help understand why this is the case: the reason for the result may lie in a matrix
of parameters learnt to fit a plausibility score on training data. Such approaches also suffer from the
out-of-vocabulary problem, where they are unable to provide results for edges involving previously
unseen nodes or edges; for example, if we add an edge SCL –flight ➛ CDG , where CDG is new to the
graph, a knowledge graph embedding will not have the entity embedding for CDG and would need
to be retrained in order to estimate the plausibility of an edge CDG –flight ➛ SCL .
https://kgbook.org 69/150
2/15/23, 5:13 PM Knowledge Graphs
(e.g., with the aforementioned rule, we can derive CDG –flight ➛ SCL from a new edge SCL –flight ➛
CDG with the unseen node CDG ).
In this section, we discuss two forms of symbolic learning: rule mining, which learns rules, and axiom
mining, which learns other forms of logical axioms.
Rule mining, in the general sense, refers to discovering meaningful patterns in the form of rules from
large collections of background knowledge. In the context of knowledge graphs, we assume a set of
positive and negative edges as given. Typically positive edges are observed edges (i.e., those given or
entailed by a knowledge graph) while negative edges are defined according to a given assumption of
completeness (discussed later). The goal of rule mining is to identify new rules that entail a high ra‐
tio of positive edges from other positive edges, but entail a low ratio of negative edges from positive
edges. The types of rules considered may vary from more simple cases, such as ?x –flight ➛ ?y ⇒ ?y
–flight ➛ ?x mentioned previously, to more complex rules, such as ?x –capital ➛ ?y –nearby ➛ ?z –type ➛
Table of Contents
Airport ⇒ ?z –type ➛ International Airport , based on observing in the graph that airports near capitals
tend to be international airports; or ⇒ ?x –domestic flight ➛ ?y , indicat‐
ing that flights within the same country denote domestic flights (as seen previously in Section 4.3.1).
Per the example inferring that airports near capital cities are international airports, rules are not as‐
sumed to hold in all cases, but rather are associated with measures of how well they conform to the
positive and negative edges. In more detail, we call the edges entailed by a rule and the set of posi‐
tive edges (not including the entailed edge itself), the positive entailments of that rule. The number of
entailments that are positive is called the support for the rule, while the ratio of a rule’s entailments
that are positive is called the confidence for the rule [Suchanek et al., 2019]. Support and confidence
indicate, respectively, the number and ratio of entailments “confirmed” to be true for the rule, where
the goal is to identify rules that have both high support and high confidence. Techniques for rule
mining in relational settings have long been explored in the context of Inductive Logic Programming
(ILP) [De Raedt, 2008]. However, knowledge graphs present novel challenges due to the scale of the
data and the frequent assumption of incomplete data (OWA), where dedicated techniques have been
proposed to address these issues [Galárraga et al., 2013].
When dealing with an incomplete knowledge graph, it is not immediately clear how to define nega‐
tive edges. A common heuristic – also used for knowledge graph embeddings – is to adopt a Partial
Completeness Assumption (PCA) [Galárraga et al., 2013], which considers the set of positive edges
to be those contained in the data graph, and the set of negative examples to be the set of all edges
x –p ➛ y′ not in the graph but where there exists a node y such that x –p ➛ y′ is in the graph.
Taking Figure 5.8, an example of a negative edge under PCA would be SCL –flight ➛ ARI (given the
presence of SCL –flight ➛ LIM ); conversely, SCL –domestic flight ➛ ARI is neither positive nor negative.
The PCA confidence measure is then the ratio of the support to the number of entailments in the pos‐
itive or negative set [Galárraga et al., 2013]. For example, the support for the rule ?x –domestic flight
➛ ?y ⇒ ?y –domestic flight ➛ ?x is 2 (since it entails IQQ –domestic flight ➛ ARI and ARI –domestic flight ➛
IQQ in the graph, which are thus positive edges), while the confidence is 2
2
= 1 (noting that SCL –
domestic flight ➛ ARI , though entailed, is neither positive nor negative, and is thus ignored by the mea‐
sure). The support for the rule ?x –flight ➛ ?y ⇒ ?y –flight ➛ ?x is analogously 4, while the confi‐
dence is 4
5
= 0.8 (noting that SCL –flight ➛ ARI is a negative edge).
The goal then, is to find rules satisfying given support and confidence thresholds. An influential rule-
mining system for graphs is AMIE [Galárraga et al., 2013, Galárraga et al., 2015], which adopts the
PCA measure of confidence, and builds rules in a top-down fashion [Suchanek et al., 2019] starting
with rule heads of the form ⇒ ?x –country ➛ ?y . For each such rule head (one for each edge label),
three types of refinements are considered, each of which adds a new edge to the body of the rule. This
new edge takes an edge label from the graph and may otherwise use fresh variables not appearing
previously in the rule, existing variables that already appear in the rule, or nodes from the graph. The
three refinements may then:
Give us feedback!
https://kgbook.org 70/150
2/15/23, 5:13 PM Knowledge Graphs
1. add an edge with one existing variable and one fresh variable; for example, refining the afore‐
mentioned rule head might give: ?z –flight ➛ ?x ⇒ ?x –country ➛ ?y ;
2. add an edge with an existing variable and a graph node; for example, refining the above rule
might give: Domestic Airport ➛ type – ?z –flight ➛ ?x ⇒ ?x –country ➛ ?y ;
3. add an edge with two existing variables; for example, refining the above rule might give:
⇒ ?x –country ➛ ?y .
These refinements can be combined arbitrarily, which gives rise to a potentially exponential search
space, where rules meeting given thresholds for support and confidence are maintained. To improve
efficiency, the search space can be pruned; for example, these three refinements always decrease
support, so if a rule does not meet the support threshold, there is no need to explore its refinements.
Further restrictions are imposed on the types of rules generated. First, only rules up to a certain fixed
size are considered. Second, a rule must be closed, meaning that each variable appears in at least two
edges of the rule, which ensures that rules are safe, meaning that each variable in the head appears
in the body; for example, the rules produced by the first and second refinements in the example are
29 The
Table of Contents
neither closed (variable y appears once) nor safe (variable y appears only notein 29
the head).
Safe rules like ?x –capital ➛ ?y –
third refinement is thus applied until a rule is closed. For further discussion ofInternational
possible optimisations
Airport are not closed as ?x appears
based on pruning and indexing, we refer to the paper [Galárraga et al., 2015].are closed is strictly stronger than the safety con
Later works have built on these techniques for mining rules from knowledge graphs. Gad-Elrab et al.
[2016] propose a method to learn non-monotonic rules – rules with negated edges in the body – in
order to capture exceptions to base rules; for example, the rule
are within the same country except when the (departure) airport is international, where the excep‐
tion is shown dotted and we use ¬ to negate an edge (representing an exception). The RuLES
system [Ho et al., 2018] – which is also capable of learning non-monotonic rules – proposes to miti‐
gate the limitations of the PCA heuristic by extending the confidence measure to consider the plausi‐
bility scores of knowledge graph embeddings for entailed edges not appearing in the graph. Where
available, explicit statements about the completeness of the knowledge graph (such as expressed in
shapes; see Section 3.1.2) can be used in lieu of PCA for identifying negative edges. Along these
lines, CARL [Pellissier Tanon et al., 2017] exploits additional knowledge about the cardinalities of
relations to refine the set of negative examples and the confidence measure for candidate rules.
Alternatively, where available, ontologies can be used to derive logically-certain negative edges un‐
der OWA through, for example, disjointness axioms. The system proposed by d’Amato et al. [d'Ama‐
to et al., 2016b, d'Amato et al., 2016a] leverages ontologically-entailed negative edges for determin‐
ing the confidence of rules generated through an evolutionary algorithm.
While the previous works involve discrete expansions of candidate rules for which a fixed confidence
scoring function is applied, another line of research is on a technique called differentiable rule min‐
ing [Rocktäschel and Riedel, 2017, Yang et al., 2017, Sadeghian et al., 2019], which allows end-to-
end learning of rules. The core idea is that the joins in rule bodies can be represented as matrix mul‐
tiplication. More specifically, we can represent the relations of an edge label p by the adjacency ma‐
trix Ap (of size ∣V ∣ × ∣V ∣) such that the value on the ith row of the j th column is 1 if there is an edge
labelled p from the ith entity to the j th entity; otherwise the value is 0. Now we can represent a join
in a rule body as matrix multiplication; for example, given ?x –domestic flight ➛ ?y ➛ country – ?z ⇒ ?x
–country ➛ ?z , we can denote the body by the matrix multiplication Adf.Ac., which gives an adjacency
matrix representing entailed country edges, where we should expect the 1’s in Adf.Ac. to be covered
by the head’s adjacency matrix Ac.. Since we are given adjacency matrices for all edge labels, we are
left to learn confidence scores for individual rules, and to learn rules (of varying length) with a
threshold confidence. Along these lines, NeuralLP [Yang et al., 2017] uses an attention mechanism to
select a variable-length sequence of edge labels for path-like rules of the form ?x –p1 ➛ y1 –p2 ➛…–pn
➛ yn –pn + 1 ➛ ?z ⇒ ?x –p ➛ ?z , for which confidences are likewise learnt. DRUM [Sadeghian et al.,
2019] also learns path-like rules, where, observing that some edge labels are more/less likely to fol‐
low others in the rules – for example, flight will not be followed by capital in the graph of Figure 5.2
Give
as the join will be empty – the system uses bidirectional recurrent neural networks (a popular tech‐ us feedback!
https://kgbook.org 71/150
2/15/23, 5:13 PM Knowledge Graphs
nique for learning over sequential data) to learn sequences of relations for rules, and their confi‐
dences. These differentiable rule mining techniques are, however, currently limited to learning path-
like rules.
More general forms of axioms beyond rules – expressed in logical languages such as DLs (see
Section 4.3.2) – can be mined from knowledge graphs. We can divide these approaches into two:
those mining specific axioms and more general axioms.
Among systems mining specific types of axioms, disjointness axioms are a popular target; for exam‐
ple, DomesticAirport ⊓ InternationalAirport ≡ ⊥ states that the two classes are disjoint by
equivalently stating that the intersection of the two classes is equivalent to the empty class, or in
simpler terms, no node can be simultaneously of type Domestic Airport and International Airport . The sys‐
tem proposed by Völker et al. [2015] extracts disjointness axioms based on (negative) association rule
mining [Agrawal et al., 1993], which finds pairs of classes where each has many instances in the
Table of Contents
knowledge graph but there are relatively few (or no) instances of both classes. Töpper et al. [2012]
rather extract disjointness for pairs of classes that have a cosine similarity below a fixed threshold.
For computing this cosine similarity, class vectors are computed using a TF–IDF analogy, where the
“document” of each class is constructed from all of its instances, and the “terms” of this document
are the properties used on the class instances (preserving multiplicities). While the previous two ap‐
proaches find disjointness constraints between named classes (e.g., city is disjoint with airport), Rizzo
et al. [2017], Rizzo et al. [2021] propose an approach that can capture disjointness constraints be‐
tween class descriptions (e.g., city without an airport nearby is disjoint with city that is the capital of a
country). The approach first clusters similar nodes of the knowledge base. Next, a terminological clus‐
ter tree is extracted, where each leaf node indicates a cluster extracted previously, and each internal
(non-leaf) node is a class definition (e.g., cities) where the left child is either a cluster having all
nodes in that class or a sub-class description (e.g., cities without airports) and the right child is either
a cluster having no nodes in that class or a disjoint-class description (e.g., non-cities with events).
Finally, candidate disjointness axioms are proposed for pairs of class descriptions in the tree that are
not entailed to have a sub-class relation.
Other systems propose methods to learn more general axioms. One of the first proposals in this direc‐
tion is the DL-FOIL system [Fanizzi et al., 2008, Rizzo et al., 2020], which is based on algorithms for
class learning (aka concept learning), whereby given a set of positive nodes and negative nodes, the
goal is to find a logical class description that divides the positive and negative sets. For example, giv‐
en { Iquique , Arica } as the positive set and { Santiago } as the negative set, we may learn a (DL) class
description ∃nearby.Airport ⊓¬(∃ capital− .⊤), denoting entities near to an airport that are not
capitals, of which all positive nodes are instances and no negative nodes are instances. Such class de‐
scriptions are learnt in an analogous manner to how aforementioned systems like AMIE learn rules,
with a refinement operator used to move from more general classes to more specific classes (and
vice-versa), a confidence scoring function, and a search strategy. Another prominent such system is
DL-Learner [Bühmann et al., 2016], which system further supports learning more general axioms
through a scoring function that uses count queries to determine what ratio of expected edges – edges
that would be entailed were the axiom true – are indeed found in the graph; for example, to score
the axiom ∃flight− .DomesticAirport ⊑ InternationalAirport over Figure 5.8, we can use
a graph query to count how many nodes have incoming flights from a domestic airport (there are 3),
and how many nodes have incoming flights from a domestic airport and are international airports
(there is 1), where the greater the difference between both counts, the weaker the evidence for the
axiom.
We now provide some abstract formal definitions for the tasks of rule mining and axiom mining over
graphs, which we generically refer to as hypothesis mining.
Give us feedback!
https://kgbook.org 72/150
2/15/23, 5:13 PM Knowledge Graphs
First we introduce hypothesis induction: a task that captures a more abstract (ideal) case for hy‐
pothesis mining. For simplicity, we focus on directed edge-labelled graphs. With a slight abuse
of notation, we may interpret a set of edges E as the graph with precisely those edges and with
no nodes or labels without edges. We may also interpret an edge e as the graph formed by {e}.
Definition 4.4; hereafter simply ⊨). Given background knowledge in the form of a knowledge
graph G (a directed edge-labelled graph, possibly extended with rules or ontologies), a set of posi‐
tive edges E + such that G does not entail any edge in E + (i.e., for all e+ ∈ E + , G
⊨ e+ ) and E +
does not contradict G (i.e., there is a model of G ∪ E + ), and a set of negative edges E − such that
G does not entail any edge in E − (i.e., for all e− ∈ E − , G
⊨ e− ), the task is to find a set of hy‐
potheses (i.e., a set of directed edge-labelled graphs) Ψ such that:
⊨ ψ for all ψ ∈ Ψ (the background knowledge does not entail any hypothesis directly);
G
G ∪ Ψ∗ ⊨ E + (the background knowledge and hypotheses together entail all positive
Table of Contents
edges);
for all e− ∈ E − , G ∪ Ψ∗
⊨ e− (the background knowledge and hypotheses together do not
entail any negative edge);
G ∪ Ψ∗ ∪ E + has a model (the background knowledge, hypotheses and positive edges taken
together do not contain a contradiction);
for all e+ ∈ E + , Ψ∗
⊨ e+ (the hypotheses alone do not entail a positive edge).
Example 5.8 — Let us assume ontological entailment ⊨ with semantic conditions Φ as defined in
Tables 4.1–4.3. Given the graph of Figure 5.8 as the background knowledge G, along with:
a set of positive edges E + = { SCL –flight ➛ ARI , SCL –domestic flight ➛ ARI }, and
a set of negative edges E − = { ARI –flight ➛ LIM , SCL –domestic flight ➛ LIM },
then a set of hypotheses Ψ = { flight –type ➛ Symmetric , domestic flight –type ➛ Symmetric } are not en‐
tailed by G, entail all positive edges in E + and no negative edges in E − when combined with G, do not
contradict G ∪ E + , and do not entail a positive edge without G. Thus Ψ satisfies the conditions for hy‐
pothesis induction.
This task represents a somewhat idealised case. Often there is no set of positive edges distinct
from the background knowledge itself. Furthermore, hypotheses not entailing a few positive
edges, or entailing a few negative edges, may still be useful. The task of hypothesis mining rather
accepts as input the background knowledge G and a set of negative edges E − (such that for all
e− ∈ E − , G
⊨ e− ), and attempts to score individual hypotheses ψ (such that G
⊨ ψ ) per their
ability to “explain” G while minimising the number of elements of E − entailed by G and ψ . We
can now abstractly define the task of hypothesis mining.
There are two scoring functions that are frequently used for σ in the literature: support and confi‐
dence.
Give us feedback!
https://kgbook.org 73/150
2/15/23, 5:13 PM Knowledge Graphs
σ + (ψ, G) := ∣{e ∈ E ∣ G′
⊨ e and G′ ∪ ψ ⊨ e}∣
where G′ denotes G with the edge e removed. Further given a set of negative edges E − , the nega‐
tive support of ψ is defined as:
σ − (ψ, G, E − ) := ∣{e− ∈ E − ∣ G ∪ ψ ⊨ e− }∣
σ + (ψ,G)
Finally, the confidence of ψ is defined as σ ± (ψ, G, E − ) := σ + (ψ,G)+σ − (ψ,G,E − ) .
We have yet to define how the set of negative edges are defined, which, in the context of a
knowledge graph G, depends on which assumption is applied:
Knowledge graphs are generally incomplete – in fact, one of the main applications of hypothesis
mining is to try to improve the completeness of the knowledge graph – and thus it would appear
unwise to assume that any edge that is not currently entailed is false/negative. We can thus rule
out CWA. Conversely, under OWA, potentially few (or no) negative edges might be entailed by
the given ontologies/rules, and thus hypotheses may end up having low negative support de‐
spite entailing many edges that do not make sense in practice. Hence the PCA can be adopted as
a heuristic to increase the number of negative edges and apply more sensible scoring of hy‐
potheses. We remark that one can adapt PCA to define negative triples by changing the subject
or predicate instead of the object.
Different implementations of hypothesis mining may consider different logical languages. Rule
mining, for example, mines hypotheses expressed either as monotonic rules (with positive
edges) or non-monotonic edges (possibly with negated edges). On the other hand, axiom mining
considers hypotheses expressed in a logical language such as Description Logics. Particular im‐
plementations may, for practical reasons, impose further syntactic restrictions on the hypotheses
generated, such as to impose thresholds on their length, on the symbols they use, or on other
structural properties (such as “closed rules” in the case of the AMIE rule mining
system [Galárraga et al., 2013]; see Section 5.4). Systems may further implement different
search strategies for hypotheses. Systems such as DL-FOIL [Fanizzi et al., 2008, Rizzo et al.,
2020], AMIE [Galárraga et al., 2013], RuLES [Ho et al., 2018], CARL [Pellissier Tanon et al.,
2017], DL-Learner [Bühmann et al., 2016], etc., propose discrete mining that recursively gener‐
ates candidate formulae through refinement/genetic operators that are then scored and checked
for threshold criteria. On the other hand, systems such as NeuralLP [Yang et al., 2017] and
DRUM [Sadeghian et al., 2019] apply differentiable mining that allows for learning (path-like)
rules and their scores in a more continuous fashion (e.g., using gradient descent). We refer to
Section 5.4 for further discussion and examples of such techniques for mining hypotheses.
https://kgbook.org 74/150
2/15/23, 5:13 PM Knowledge Graphs
One approach for creating and enriching knowledge graphs is to solicit direct contributions from hu‐
man editors. Such editors may be found in-house (e.g., employees of the tourist board), using crowd-
sourcing platforms, through feedback mechanisms (e.g., tourists adding comments on attractions),
through collaborative-editing platforms (e.g., an attractions wiki open to public edits), etc. Though
human involvement incurs high costs [Paulheim, 2018], some prominent knowledge graphs have
been primarily based on direct contributions from human editors [Vrandečić and Krötzsch, 2014, He
et al., 2016]. Depending on how the contributions are solicited, however, the approach has a number
of key drawbacks, due primarily to human error [Pellissier Tanon et al., 2016],
disagreement [Yasseri et al., 2012], bias [Janowicz et al., 2018], vandalism [Heindorf et al., 2016],
etc. Successful collaborative creation further raises challenges concerning licensing, tooling, and cul‐
ture [Pellissier Tanon et al., 2016]. Humans are sometimes rather employed to verify and curate ad‐
ditions to a knowledge graph extracted by other means [Pellissier Tanon et al., 2016] (through, e.g.,
video games with a purpose [Jurgens and Navigli, 2014]), to define high-quality mappings from oth‐
er sources [Das et al., 2012], to define appropriate high-level schema [Keet, 2018, Labra Gayo et al.,
2018], and so forth.
Text corpora – such as sourced from newspapers, books, scientific articles, social media, emails, web
crawls, etc. – are an abundant source of rich information [Hellmann et al., 2013, Rospocher et al.,
2016]. However, extracting such information with high precision and recall for the purposes of cre‐
ating or enriching a knowledge graph is a non-trivial challenge. To address this, techniques from
Natural Language Processing (NLP) [Maynard et al., 2016, Jurafsky and Martin, 2019] and
Information Extraction (IE) [Weikum and Theobald, 2010, Grishman, 2012, Martínez-Rodríguez et
al., 2020] can be applied. Though processes vary considerably across text extraction frameworks, in
Figure 6.1 we illustrate four core tasks for text extraction on a sample sentence. We will discuss
these tasks in turn.
Figure 6.1: Text extraction example; dashed nodes are new to the knowledge graph
Give us feedback!
https://kgbook.org 75/150
2/15/23, 5:13 PM Knowledge Graphs
6.2.1 Pre-processing
The pre-processing task may involve applying various techniques to the input text, where Figure 6.1
illustrates Tokenisation, which parses the text into atomic terms and symbols. Other pre-processing
tasks applied to a text corpus may include: Part-of-Speech (POS) tagging [Maynard et al., 2016,
Jurafsky and Martin, 2019] to identify terms representing verbs, nouns, adjectives, etc.; Dependency
Parsing, which extracts a grammatical tree structure for a sentence where leaf nodes indicate individ‐
ual words that together form phrases (e.g., noun phrases, verb phrases) and eventually clauses and
sentences [Maynard et al., 2016, Jurafsky and Martin, 2019]; and Word Sense Disambiguation
(WSD) [Navigli, 2009] to identify the meaning (aka sense) in which a word is used, linking words
with a lexicon of senses (e.g., WordNet [Miller and Fellbaum, 2007] or BabelNet [Navigli and
Ponzetto, 2012]), where, for instance, the term flights may be linked with the WordNet sense “an in-
stance of travelling by air” rather than “a stairway between one floor and the next”. The appropriate type of
pre-processing to apply often depends on the requirements of later tasks in the pipeline.
The NER task identifies mentions of named entities in a text [Nadeau and Sekine, 2007, Ratinov and
Roth, 2009], typically targetting mentions of people, organisations, locations, and potentially other
types [Ling and Weld, 2012, Nakashole et al., 2013, Yogatama et al., 2015]. A variety of NER tech‐
niques exist, with many modern approaches based on learning frameworks that leverage lexical fea‐
tures (e.g., POS tags, dependency parse trees, etc.) and gazetteers (e.g., lists of common first names,
last names, countries, prominent businesses, etc.). Supervised methods [Bikel et al., 1999, Finkel et
al., 2005, Lample et al., 2016] require manually labelling all entity mentions in a training corpus,
whereas bootstrapping-based approaches [Collins and Singer, 1999, Etzioni et al., 2004, Nakashole et
al., 2013, Gupta and Manning, 2014] rather require a small set of seed examples of entity mentions
from which patterns can be learnt and applied to unlabelled text. Distant supervision [Ling and Weld,
2012, Ren et al., 2015, Yogatama et al., 2015] uses known entities in a knowledge graph as seed ex‐
amples through which similar entities can be detected. Aside from learning-based frameworks, tradi‐
tional approaches based on manually-crafted rules [Kluegl et al., 2009, Chiticariu et al., 2018] are
still sometimes used due to their more controllable and predictable behaviour [Chiticariu et al.,
2013]. The named entities identified by NER may be used to generate new candidate nodes for the
knowledge graph (known as emerging entities, shown dashed in Figure 6.1), or may be linked to exist‐
ing nodes per the Entity Linking task described in the following.
The EL task associates mentions of entities in a text with the existing nodes of a target knowledge
graph, which may be the nucleus of a knowledge graph under creation, or an external knowledge
graph [Wu et al., 2018]. In Figure 6.1, we assume that the nodes Santiago and Easter Island already
exist in the knowledge graph (possibly extracted from other sources). EL may then link the given
mentions to these nodes. The EL task presents two main challenges. First, there may be multiple
ways to mention the same entity, as in the case of Rapa Nui and Easter Island ; if we created a node
Rapa Nui to represent that mention, we would split the information available under both mentions
across different nodes, where it is thus important for the target knowledge graph to capture the vari‐
ous aliases and multilingual labels by which one can refer to an entity [Moro et al., 2014]. Second,
the same mention in different contexts can refer to distinct entities; for instance, Santiago can refer to
cities in Chile, Cuba, Spain, amongst others. The EL task thus considers a disambiguation phase where‐
in mentions are associated to candidate nodes in the knowledge graph, the candidates are ranked,
and the most likely node being mentioned is chosen [Wu et al., 2018]. Context can be used in this
phase; for example, if Easter Island is a likely candidate for the corresponding mention alongside
Santiago , we may boost the probability that this mention refers to the Chilean capital as both candi‐
dates are located in Chile. Other heuristics for disambiguation consider a prior probability, where for
example, Santiago most often refers to the Chilean capital (being, e.g., the largest city with that
name); centrality measures on the knowledge graph can be used for such purposes [Wu et al., 2018].
Give us feedback!
https://kgbook.org 76/150
2/15/23, 5:13 PM Knowledge Graphs
6.2.4 Relation Extraction (RE)
The RE task extracts relations between entities in the text [Zhou et al., 2005, Bach and Badaskar,
2007]. The simplest case is that of extracting binary relations in a closed setting wherein a fixed set of
relation types are considered. While traditional approaches often relied on manually-crafted
patterns [Hearst, 1992], modern approaches rather tend to use learning-based frameworks [Roller et
al., 2018], including supervised methods over manually-labelled examples [Bunescu and Mooney,
2005, Zhou et al., 2005]. Other learning-based approaches again use bootstrapping [Etzioni et al.,
2004, Bunescu and Mooney, 2007] and distant supervision [Mintz et al., 2009, Riedel et al., 2010,
Hoffmann et al., 2011, Surdeanu et al., 2012, Xu et al., 2013, Smirnova and Cudré-Mauroux, 2019]
to forgo the need for manual labelling; the former requires a subset of manually-labelled seed exam‐
ples, while the latter finds sentences in a large corpus of text mentioning pairs of entities with a
known relation/edge, which are used to learn patterns for that relation. Binary RE can also be ap‐
plied using unsupervised methods in an open setting – often referred to as Open Information
Extraction (OIE) [Banko et al., 2007, Etzioni et al., 2011, Fader et al., 2011, Mausam et al., 2012,
Mausam, 2016, Mitchell et al., 2018] – whereby the set of target relations is not pre-defined but
Table of Contents
rather extracted from text based on, for example, dependency parse trees from which relations are
taken.
A variety of RE methods have been proposed to extract n-ary relations that capture further context
for how entities are related. In Figure 6.1, we see how an n-ary relation captures additional temporal
context, denoting when Rapa Nui was named a World Heritage site; in this case, an anonymous node
is created to represent the higher-arity relation in the directed-labelled graph. Various methods for n
-ary RE are based on frame semantics [Fillmore, 1976], which, for a given verb (e.g., “named”), cap‐
tures the entities involved and how they may be interrelated. Resources such as FrameNet [Baker et
al., 1998] then define frames for words, which, for example, may identify that the semantic frame
for “named” includes a speaker (the person naming something), an entity (the thing named) and a
name. Optional frame elements are an explanation, a purpose, a place, a time, etc., that may add con‐
text to the relation. Other RE methods are rather based on Discourse Representation Theory
(DRT) [Kamp, 1981], which considers a logical representation of text based on existential events.
Under this theory, for example, the naming of Easter Island as a World Heritage Site is considered to
be an (existential) event where Easter Island is the patient (the entity affected), leading to the logical
(neo-Davidsonian) formula:
Such a formula is analogous to reification, as discussed previously in Section 3.3, where e is an exis‐
tential term that refers to the n-ary relation being extracted.
Finally, while relations extracted in a closed setting are typically mapped directly to a knowledge
graph, relations that are extracted in an open setting may need to be aligned with the knowledge
graph; for example, if an OIE process extracts a binary relation Santiago –has flights to ➛ Easter Island , it
may be the case that the knowledge graph does not have other edges labelled has flights to, where
alignment may rather map such a relation to the edge Santiago –flight ➛ Easter Island assuming flight is
used in the knowledge graph. A variety of methods have been applied for performing such align‐
ments, including mappings [Corcoglioniti et al., 2016, Gangemi et al., 2017] and rules [Rouces et al.,
2015] for aligning n-ary relations; distributional and dependency-based similarities [Moro and
Navigli, 2013], association rule mining [Dutta et al., 2014], Markov clustering [Dutta et al., 2015]
and linguistic techniques [Martínez-Rodríguez et al., 2018] for aligning OIE relations; amongst
others.
Having presented the four main tasks for building knowledge graphs from text, it is important to
note that frameworks do not always follow this particular sequence of tasks. A common trend, for ex‐
ample, is to combine interdependent tasks, jointly performing WSD and EL [Moro et al., 2014], or
NER and EL [Luo et al., 2015, Nguyen et al., 2016], or NER and RE [Ren et al., 2017, Zheng et al.,
Give
2017], etc., in order to mutually improve the performance of multiple tasks. For further details on us feedback!
https://kgbook.org 77/150
2/15/23, 5:13 PM Knowledge Graphs
extracting knowledge graphs from text we refer to the book by Maynard et al. [2016] and the recent
survey by Martínez-Rodríguez et al. [2020].
The Web was founded on interlinking markup documents wherein markers (aka tags) are used to sepa‐
rate elements of the document (typically for formatting purposes). Most documents on the Web use
the HyperText Markup Language (HTML). Figure 6.2 presents an example HTML webpage about
World Heritage Sites in Chile. Other formats of markup include Wikitext used by Wikipedia, TeX for
typesetting, Markdown used by Content Management Systems, etc. One approach for extracting in‐
formation from markup documents – in order to create and/or enrich a knowledge graph – is to strip
the markers (e.g., HTML tags), leaving only plain text upon which the techniques from the previous
section can be applied. However, markup can be useful for extraction purposes, where variations of
the aforementioned tasks for text extraction have been adapted to exploit such markup [Lu et al.,
2013, Lockard et al., 2018, Martínez-Rodríguez et al., 2020]. We can divide extraction techniques for
Table of Contents
markup documents into three main categories: general approaches that work independently of the
markup used in a particular format, often based on wrappers that map elements of the document to
the output; focussed approaches that target specific forms of markup in a document, most typically
web tables (but sometimes also lists, links, etc.); and form-based approaches that extract the data un‐
derlying a webpage, per the notion of the Deep Web. These approaches can often benefit from the
regularities shared by webpages of a given website; for example, intuitively speaking, while the web‐
page of Figure 6.2 is about Chile, we will likely find pages for other countries following the same
structure on the same website.
<html>
<head><title>UNESCO World Heritage Sites</title></head>
<body>
<h1>World Heritage Sites</h1> ? UNESCO World Heritage Sites ×
<h2>Chile</h2>
<p>Chile has 6 UNESCO World Heritage Sites.</p>
World Heritage Sites
<table border="1"> Chile
<tr><th>Place</th><th>Year</th><th>Criteria</th></t Chile has 6 UNESCO World Heritage Sites.
<tr><td>Rapa Nui</td><td>1995</td>
<td rowspan="6">Cultural</td></tr> Place Year Criteria
<tr><td>Churches of Chiloé</td><td>2000</td></tr> Rapa Nui 1995
<tr><td>Historical Valparaíso</td><td>2003</td></tr Churches of Chiloé 2000
<tr><td>Saltpeter Works</td><td>2005</td></tr> Historical Valparaíso 2003
Cultural
<tr><td>Sewell Mining Town</td><td>2006</td></tr> Saltpeter Works 2005
<tr><td>Qhapaq Ñan</td><td>2014</td></tr> Sewell Mining Town 2006
</table> Qhapaq Ñan 2014
</body>
</html>
Figure 6.2: Example markup document (HTML) with source-code (left) and formatted document (right)
Many general approaches are based on wrappers that locate and extract the useful information direct‐
ly from the markup document. While the traditional approach was to define such wrappers manually
– a task for which a variety of declarative languages and tools have been defined – such approaches
are brittle to changes in a website’s layout [Ferrara et al., 2014]. Hence other approaches allow for
(semi-)automatically inducing wrappers [Flesca et al., 2004]. A modern such approach – used to en‐
rich knowledge graphs in systems such as LODIE [Gentile et al., 2014] – is to apply distant supervi‐
sion, whereby EL is used to identify and link entities in the webpage to nodes in the knowledge
graph such that paths in the markup that connect pairs of nodes for known edges can be extracted,
ranked, and applied to other examples. Taking Figure 6.2, for example, distant supervision may link
Rapa Nui and World Heritage Sites to the nodes Easter Island and World Heritage Site in the knowledge
graph using EL, and given the edge Easter Island –named ➛ World Heritage Site in the knowledge graph
(extracted per Figure 6.1), identify the candidate path (x, td [1]− ⋅ tr − ⋅ table − ⋅ h1 , y) as reflecting edges
of the form x –named ➛ y , where t[n] indicates the nth child of tag t, t− its inverse, and t1 ⋅ t2 con‐
Give
catenation. Finally, paths with high confidence (e.g., ones “witnessed” by many known edges in the us feedback!
https://kgbook.org 78/150
2/15/23, 5:13 PM Knowledge Graphs
knowledge graph) can then be used to extract novel edges, such as Qhapaq Ñan –named ➛
World Heritage Site , both on this page and on related pages of the website with similar structure (e.g.,
for other countries).
Other approaches target specific types of markup, most commonly web tables embedded in HTML
webpages. However, web tables are designed to enhance human rather than machine readability.
Many web tables are used for layout and page structure (e.g., navigation bars). Those that contain
data may follow different formats, such as relational tables, listings, attribute-value tables, and ma‐
trices [Cafarella et al., 2008, Crestan and Pantel, 2011]. A first step is to classify tables to find ones
appropriate for the given extraction mechanism(s) [Crestan and Pantel, 2011, Eberius et al., 2015].
Next, web tables may contain column spans, row spans, inner tables, or may be split vertically to im‐
prove human aesthetics. Table normalisation merges split tables, un-nests tables, transposes tables,
etc. [Pivk et al., 2007, Cafarella et al., 2008, Crestan and Pantel, 2011, Deng et al., 2013, Ermilov
and Ngonga Ngomo, 2016, Lehmberg et al., 2016]. Some approaches then identify the table protago‐
Table of Contents
nist [Crestan and Pantel, 2011, Muñoz et al., 2014] – the main entity that the table describes – often
found elsewhere in the webpages; for example, though not mentioned by the table of Figure 6.1,
World Heritage Sites is its protagonist. Finally, extraction processes may associate cells with
entities [Limaye et al., 2010, Mulwad et al., 2013], columns with types [Deng et al., 2013, Limaye et
al., 2010, Mulwad et al., 2013], and column pairs with relations [Limaye et al., 2010, Muñoz et al.,
2014]. When enriching knowledge graphs, recent approaches apply distant supervision, linking cells
to knowledge graph nodes in order to generate candidates for type and relation extraction [Limaye
et al., 2010, Mulwad et al., 2013, Muñoz et al., 2014]. Statistical distributions can also help to link
numerical columns [Neumaier et al., 2016]. Specialised table extraction frameworks have also been
proposed for specific websites, where prominent knowledge graphs, such as DBpedia [Lehmann et
al., 2015] and YAGO [Suchanek et al., 2008] focus on extraction from info-box tables in Wikipedia.
The Deep Web presents a rich source of information accessible only through searches on web forms,
thus requiring Deep Web crawling techniques to access [Madhavan et al., 2008]. Systems have been
proposed to extract knowledge graphs from Deep Web sources [Geller et al., 2008, Lehmann et al.,
2012, Collarana et al., 2016]. Approaches typically attempt to generate sensible form inputs – which
may be based on a user query or generated from reference knowledge – and then extract data from
the generated responses (markup documents) using the aforementioned techniques [Geller et al.,
2008, Lehmann et al., 2012, Collarana et al., 2016].
Much of the legacy data available within organisations and on the Web is represented in structured
formats, primarily tables – in the form of relational databases, CSV files, etc. – but also tree-struc‐
tured formats such as JSON, XML etc. Unlike text and markup documents, structured sources can of‐
ten be mapped to knowledge graphs whereby the structure is (precisely) transformed according to a
mapping rather than (imprecisely) extracted. The mapping process involves two steps: 1) create a
mapping from the source to a graph, and 2) use the mapping in order to materialise the source data
as a graph or to virtualise the source (creating a graph view over the legacy data).
Tabular sources of data are prevalent; for example, the structured content underlying many organisa‐
tions and websites are housed in relational databases. In Figure 6.3 we present an example of a rela‐
tional database instance that we wish to integrate into our knowledge graph. There are then two ap‐
proaches for mapping content from tables to knowledge graphs: a direct mapping, and a custom
mapping. Give us feedback!
https://kgbook.org 79/150
2/15/23, 5:13 PM Knowledge Graphs
Report
crime claimant station date Claimant
Pickpocketing XY12SDA Viña del Mar 2019-04-12 id name country
Assault AB9123N Arica 2019-04-12 XY12SDA John Smith U.S.
Pickpocketing XY12SDA Rapa Nui 2019-04-12 AB9123N Joan Dubois France
Fraud FI92HAS Arica 2019-04-13 XI92HAS Jorge Hernández Chile
Figure 6.3: Relational database instance with two tables describing crime data
Table of Contents
Figure 6.4: Direct mapping result for the first rows of both tables in Figure 6.3
A direct mapping automatically generates a graph from a table. We present in Figure 6.4 the result
of a standard direct mapping [Arenas et al., 2012], which creates an edge x –y ➛ z for each (non-
header, non-empty, non-null) cell of the table, such that x represents the row of the cell, y the col‐
umn name of the cell, and z the value of the cell. In particular, x typically encodes the values of
the primary key for a row (e.g., Claimant.id); otherwise, if no primary key is defined (e.g., per the
Report table), x can be an anonymous node or a node based on the row number. The node x and
edge label y further encode the name of the table to avoid clashes across tables that have the same
column names used with different meanings. For each row x , we may add a type edge based on the
name of its table. The value z may be mapped to datatype values in the corresponding graph model
based on the source domain (e.g., a value in an SQL column of type Date can be mapped to
xsd:date in the RDF data model). If the value is null (or empty), typically the corresponding edge
will be omitted.30 With respect to Figure 6.4, we highlight the difference note between the nodes
30 One might consider representing nulls
Claimant-XY12SDA and XY12SDA , where the former denotes the row (or entity)null s in SQL
identified bycan
thebe used to mean that there is n
latter
istential semantics of such nodes (e.g., in RDF).
primary key value. In case of a foreign key between two tables – such as Report.claimant referenc‐
ing Claimant.id – we can link, for example, to Claimant-XY12SDA rather than XY12SDA , where the
former node also has the name and country of the claimant. A direct mapping along these lines has
been standardised for mapping relational databases to RDF [Arenas et al., 2012], where Stoica et al.
[2019] have recently proposed an analogous direct mapping for property graphs. Another direct
mapping has been defined for CSV and other tabular data [Tandy et al., 2015] that further allows for
specifying column names, primary/foreign keys, and data types – which are often missing in such
data formats – as part of the mapping itself.
Although a direct mapping can be applied automatically on tabular sources of data and preserve the
information of the original source – i.e., allowing a deterministic inverse mapping that reconstructs
the tabular source from the output graph [Sequeda et al., 2012] – in many cases it is desirable to
customise a mapping, such as to align edge labels or nodes with a knowledge graph under enrich‐
ment, etc. Along these lines, declarative mapping languages allow for manually defining custom
mappings from tabular sources to graphs. A standard language along these lines is the RDB2RDF
Mapping Language (R2RML) [Das et al., 2012], which allows for mapping from individual rows of a
table to one or more custom edges, with nodes and edges defined either as constants, as individual
cell values, or using templates that concatenate multiple cell values from a row and static substrings
into a single term; for example, a template {id}-{country} may produce nodes such as
XY12SDA-U.S. from the Claimant table. In case that the desired output edges cannot be defined from
a single row, R2RML allows for (SQL) queries to generate tables from which edges can be extracted
where, for example, edges such as U.S. –crimes ➛ 2 can be generated by defining the mapping with
respect to a query that joins the Report and Claimant tables on claimant=id, grouping by
country, and applying a count for each country group. A mapping can then be defined on the re‐
sults table such that the source node denotes the value of country, the edge label is the constant
crimes, and the target node is the count value. An analogous standard also exists for mapping CSV
Give us feedback!
https://kgbook.org 80/150
2/15/23, 5:13 PM Knowledge Graphs
and other tabular data to RDF graphs, again allowing keys, column names, and datatypes to be cho‐
sen as part of the mapping [Tennison and Kellogg, 2015].
Once the mappings have been defined, one option is to use them to materialise graph data following
an Extract-Transform-Load (ETL) approach, whereby the tabular data are transformed and explicitly
serialised as graph data using the mapping. A second option is to use virtualisation through a Query
Rewriting (QR) approach, whereby queries on the graph (using, e.g., SPARQL, Cypher, etc.) are trans‐
lated to queries over the tabular data (typically using SQL). Comparing these two options, ETL al‐
lows the graph data to be used as if they were any other data in the knowledge graph. However, ETL
requires updates to the underlying tabular data to be explicitly propagated to the knowledge graph,
whereas a QR approach only maintains one copy of data to be updated. The area of Ontology-Based
Data Access (OBDA) [Xiao et al., 2018] is concerned with QR approaches that support ontological en‐
tailments as seen in Chapter 4. Although most QR approaches only support non-recursive entail‐
ments expressible as a single (non-recursive) query, some QR approaches support recursive entail‐
ments through rewritings to recursive queries [Sequeda et al., 2014].
Table of Contents
A number of popular data formats are based on trees, including XML and JSON. While one could
imagine – leaving aside issues such as the ordering of children in a tree – a trivial direct mapping
from trees to graphs by simply creating edges of the form x –child ➛ y for each node y that is a child
of x in the source tree, such an approach is not typically used, as it represents the literal structure of
the source data. Instead, the content of tree-structured data can be more naturally represented as a
graph using a custom mapping. Along these lines, the GRDLL standard [Connolly, 2007] allows for
mapping from XML to (RDF) graphs, while languages such as RML allow for mapping from a variety
of formats, including XML and JSON, to (RDF) graphs [Dimou et al., 2014]. In contrast, hybrid query
languages such as XSPARQL [Bischof et al., 2012] allow for querying XML and RDF in unison, thus
supporting both materialisation and virtualisation of graphs over tree-structured sources of legacy
data.
We may also leverage existing knowledge graphs in order to construct or enrich another knowledge
graph. For example, a large number of points of interest for the Chilean tourist board may be avail‐
able in existing knowledge graphs such as BabelNet [Navigli and Ponzetto, 2012],
DBpedia [Lehmann et al., 2015], LinkedGeoData [Stadler et al., 2012], Wikidata [Vrandečić and
Krötzsch, 2014], YAGO [Hoffart et al., 2011], etc. However, not all entities and/or relations may be
of interest. A standard option to extract a relevant sub-graph of data is to use construct queries that
generate graphs as output [Neumaier and Polleres, 2019]. Entity and schema alignment between the
knowledge graphs may be further necessary to better integrate (parts of) external knowledge graphs,
using linking tools for graphsexternal identifiers [Pellissier Tanon et al., 2016], or indeed may be
done manually [Pellissier Tanon et al., 2016]. For instance, Wikidata [Vrandečić and Krötzsch,
2014] uses Freebase [Bollacker et al., 2007b, Pellissier Tanon et al., 2016] as a source; Gottschalk
and Demidova [2018] extract an event-centric knowledge graph from Wikidata, DBpedia and YAGO;
while Neumaier and Polleres [2019] construct a spatio-temporal knowledge graph from Geonames,
Wikidata, and PeriodO [Golden and Shaw, 2016] (as well as tabular data).
The discussion thus far has focussed on extracting data from external sources in order to create and
enrich a knowledge graph. In this section, we discuss some of the principal methods for generating a
schema based on external sources of data, including human knowledge. For discussion on extracting
a schema from the knowledge graph itself, we refer back to Section 3.1.3. In general, much of the
work in this area has focussed on the creation of ontologies using either ontology engineering
methodologies, and/or ontology learning. We discuss these two approaches in turn.
Give us feedback!
https://kgbook.org 81/150
2/15/23, 5:13 PM Knowledge Graphs
6.5.1 Ontology engineering
Ontology engineering refers to the development and application of methodologies for building on‐
tologies, proposing principled processes by which better quality ontologies can be constructed and
maintained with less effort. Early methodologies [Grüninger and Fox, 1995a, Fernández et al., 1997,
Noy and McGuinness, 2001] were often based on a waterfall-like process, where requirements and
conceptualisation were fixed before starting to define the ontology, using, for example, an ontology
engineering tool [Gómez-Pérez et al., 2006, Keet, 2018, Kendall and McGuinness, 2019]. However,
for situations involving large or ever-evolving ontologies, more iterative and agile ways of building
and maintaining ontologies have been proposed.
DILIGENT [Pinto et al., 2009] was an early example of an agile methodology, proposing a complete
process for ontology life-cycle management and knowledge evolution, as well as separating local
changes (local views on knowledge) from global updates of the core part of the ontology, using a re‐
view process to authorise the propagation of changes from the local to the global level. This method‐
ology is similar to how, for instance, the large clinical reference terminology SNOMED CT [IHTSDO,
Table of Contents
2019] (also available as an ontology) is maintained and evolved, where the (international) core ter‐
minology is maintained based on global requirements, while national or local extensions to SNOMED
CT are maintained based on local requirements. A group of authors then decides which national or
local extensions to propagate to the core terminology. More modern agile methodologies include eX‐
treme Design (XD) [Presutti et al., 2009, Blomqvist et al., 2016], Modular Ontology Modelling
(MOM) [Krisnadhi and Hitzler, 2016b, Hitzler and Krisnadhi, 2018], Simplified Agile Methodology
for Ontology Development (SAMOD) [Peroni, 2016], and more besides. Such methodologies typical‐
ly include two key elements: ontology requirements and (more recently) ontology design patterns.
Ontology requirements specify the intended task of the resulting ontology, or of the knowledge
graph itself in conjunction with the new ontology. A common way to express ontology requirements
is through Competency Questions (CQ) [Grüninger and Fox, 1995b], which are natural language ques‐
tions illustrating the typical information needs that one would require the ontology (or the knowl‐
edge graph) to respond to. Such CQs can then be complemented with additional restrictions, and rea‐
soning requirements, in case that the ontology should also contain restrictions and general axioms
for inferring new knowledge or checking data consistency. A common way of testing ontologies (or
knowledge graphs based on them) is then to formalise the CQs as queries over some test set of data,
and make sure the expected results are entailed [Blomqvist et al., 2012, Keet and Ławrynowicz,
2016]. We may, for example, consider the CQ “What are all the events happening in Santiago?”, which
can be represented as a graph query Event ➛ type – ?event –location ➛ Santiago . Taking the data graph of
Figure 2.1 and the axioms of Figure 3.2, we can check to see if the expected result EID15 is entailed
by the ontology and the data, and since it is not, we may consider expanding the axioms to assert
that location –type ➛ Transitive .
Ontology Design Patterns (ODPs) are another common feature of modern methodologies [Gangemi,
2005, Blomqvist and Sandkuhl, 2005], specifying generalisable ontology modelling patterns that can
be used as inspiration for modelling similar patterns, as modelling templates [Egaña et al., 2008,
Skjæveland et al., 2018], or as directly reusable components [Daga et al., 2008, Shimizu et al.,
2019]. Several pattern libraries have been made available online, ranging from carefully curated
ones [Aranguren et al., 2008, Shimizu et al., 2019] to open and community moderated ones [Daga et
al., 2008]. As an example, to model events in our scenario, we may adopt the Core Event ontology
pattern proposed by Krisnadhi and Hitzler [2016a], which specifies a spatio-temporal extent, sub-
events, and participants of an event, along with competency questions, formal definitions, etc., to
support this pattern.
The previous methodologies outline methods by which ontologies can be built and maintained man‐
ually. Ontology learning, in contrast, can be used to (semi-)automatically extract information from
text that is useful for the ontology engineering process [Buitelaar et al., 2005, Cimiano, 2006]. Early
methods focussed on extracting terminology from text that may represent the relevant domain’s
Give us feedback!
https://kgbook.org 82/150
2/15/23, 5:13 PM Knowledge Graphs
classes; for example, from a collection of text documents about tourism, a terminology extraction
tool – using measures of unithood that determine how cohesive an n-gram is as a unitary phrase, and
termhood that determine how relevant the phrase is to a domain [Martínez-Rodríguez et al., 2018] –
may identify n-grams such as “visitor visa”, “World Heritage Site”, “off-peak rate”, etc., as terminology of
particular importance to the tourist domain that thus may merit inclusion in such an ontology.
Ontological axioms may also be extracted from text. A common target is to extract sub-class axioms
from text, leveraging patterns based on modifying nouns and adjectives that incrementally specialise
concepts (e.g., extracting Visitor Visa –subc. of ➛ Visa from the noun phrase “visitor visa” and isolated
appearances of “visa” elsewhere), or using Hearst patterns [Hearst, 1992] (e.g., extracting
Off-Peak Rate –subc. of ➛ Discount from “many discounts, such as off-peak rates, are available” based on the
pattern “X, such as Y”). Textual definitions can also be harvested from large texts to extract hyper‐
nym relations and induce a taxonomy from scratch [Velardi et al., 2013]. More recent works aim to
extract more expressive axioms from text, including disjointness axioms [Völker et al., 2015]; and
axioms involving the union and intersection of classes, along with existential, universal, and quali‐
fied-cardinality restrictions [Petrucci et al., 2016]. The results of an ontology learning process can
then serve as input to a more general ontology engineering methodology, allowing us to validate the
Table of Contents
Chapter 7
Quality Assessment
Independently of the (kinds of) source(s) from which a knowledge graph is created, the resulting ini‐
tial knowledge graph will usually be incomplete, and will often contain duplicate, contradictory or
even incorrect statements, especially when taken from multiple sources. After the initial creation and
enrichment of a knowledge graph from external sources, a crucial step is thus to assess the quality of
the resulting knowledge graph. By quality, we here refer to fitness for purpose. Quality assessment
then helps to ascertain for which purposes a knowledge graph can be reliably used. Take, for in‐
stance, the sample of an initial knowledge graph created by the tourist board shown in Figure 7.1. Is
this knowledge graph of good quality? Does it exhibit issues that might limit the applications for
which it is fit for purpose? Can we define and detect such issues? These questions are crucial to ad‐
dress before the knowledge graph is deployed, but they are also challenging to address in a general
way.
Figure 7.1: A newly created knowledge graph about events and their venues
This chapter discusses (sometimes overlapping) quality dimensions that capture qualitative aspects of
the multifaceted notion of data quality; some of these dimensions apply more generally to
databases [Batini et al., 2015], while others are more specific to knowledge graphs [Zaveri et al.,
Give
2016]. We further discuss quality metrics that provide ways to measure quantitative aspects of these us feedback!
https://kgbook.org 83/150
2/15/23, 5:13 PM Knowledge Graphs
dimensions. We group dimensions and metrics in a manner inspired by Batini and Scannapieco
[2016].
7.1 Accuracy
Accuracy refers to the extent to which entities and relations – encoded by nodes and edges in the
graph – correctly represent real-life phenomena. Accuracy can be divided into three dimensions: syn‐
tactic accuracy, semantic accuracy, and timeliness.
Syntactic accuracy is the degree to which the data are accurate with respect to the grammatical rules
defined for the domain and/or data model. A prevalent example of syntactic inaccuracy occurs with
datatype nodes, which may be incompatible with a defined range or be malformed. For example, as‐
suming that a property start is defined with the range xsd:dateTime, the value March 29, 2019 in
Table of Contents
Figure 7.1 would be incompatible with the defined range, while a value
“March 29, 2019, 20:00”^^xsd:dateTime would be malformed (a value such as
“2019-03-22T20:00:00”^^xsd:dateTime is rather expected). A corresponding metric for syntactic accuracy is
the ratio between the number of invalid values of a given property and the total number of values
for the same property [Zaveri et al., 2016]. Such forms of syntactic accuracy can typically be as‐
sessed using validation tools [Fürber and Hepp, 2011, Hogan et al., 2010].
Semantic accuracy is the degree to which data values correctly represent real-world phenomena,
which may be affected by imprecise extraction results, untrustworthy sources, vandalism, etc. For in‐
stance, in Figure 7.1, the start of the EID15 event comes after the end of the event, possibly due to a
typo in the year. While such a case could potentially be identified using, for example, shape-based
validation, other cases might be more difficult to detect; for example, if we were to accidentally (and
incorrectly) swap the venues for EID15 and EID17 , there might be no indication whatsoever in the
knowledge graph that the venues are incorrect, even if we have additional
schemata/ontologies/rules available. Assessing the level of semantic inaccuracy is challenging. While
one option is to apply manual verification, an automatic option may be to check the stated relation
against several sources [Lei et al., 2007, Esteves et al., 2018]. An alternative is to validate the quality
of the processes used to generate the knowledge graph, based on measures such as precision, possi‐
bly with the help of human experts or gold standards [Martínez-Rodríguez et al., 2020].
7.1.3 Timeliness
Timeliness is the degree to which the knowledge graph is kept up-to-date with the real world
state [Käfer et al., 2013]. A knowledge graph may be semantically accurate now, but may quickly
become inaccurate (outdated) if no procedures are in place to keep it up-to-date in a timely manner.
Considering Figure 7.1, the events appear to be from years ago, and if not updated, then the knowl‐
edge graph will not be suitable for applications that wish to recommend upcoming events to users.
Additionally, the meaning of some values in the graph, such as Next Tuesday or Next Thursday (which
may have been extracted from the text of a news article, for example), will change over time, and
become semantically inaccurate in the future. Similarly, the age of Santiago will quickly become out‐
dated, where instead representing the year that the city was founded would facilitate timeliness.
Timeliness can be assessed based on how frequently the knowledge graph is updated with respect to
underlying sources [Käfer et al., 2013, Rula et al., 2014], which can be done using temporal annota‐
tions of changes in the knowledge graph [Rula et al., 2012, Rula et al., 2019], as well as contextual
representations that capture the temporal validity of data (see Section 3.3).
https://kgbook.org 84/150
2/15/23, 5:13 PM Knowledge Graphs
Coverage refers to avoiding the omission of domain-relevant elements, which otherwise may yield
incomplete query results or entailments, biased models, etc.
7.2.1 Completeness
Completeness refers to the degree to which all required information is present in a particular dataset.
Completeness comprises the following aspects: (i) schema completeness refers to the degree to which
the classes and properties of a schema are represented in the data graph, (ii) property completeness
refers to the ratio of missing values for a specific property, (iii) population completeness refers to the
percentage of all real-world entities of a particular type that are represented in the datasets, and (iv)
linkability completeness refers to the degree to which instances in the data set are interlinked. Taking
some examples from Figure 7.1, the lack of information about the fare for EID15 might be seen as a
form of property incompleteness, while missing events held in Chile around the same time might
lead to population incompleteness. Measuring completeness is non-trivial as it assumes knowledge of
a hypothetical ideal knowledge graph [Darari et al., 2018] that contains all the elements that the
knowledge graph in question should have. Concrete strategies may involve comparison with gold
Table of Contents
standards that provide samples of the ideal knowledge graph (possibly based on completeness state‐
ments [Darari et al., 2018]), or measuring the recall of extraction methods from complete
sources [Martínez-Rodríguez et al., 2020].
7.2.2 Representativeness
Representativeness is a related dimension that, instead of focusing on the ratio of domain-relevant ele‐
ments that are missing, rather focuses on assessing high-level biases in what is included/excluded
from the knowledge graph [Baeza-Yates, 2018]. As such, this dimension assumes that the knowledge
graph is incomplete – i.e., that it is a sample of the ideal knowledge graph – and asks how biased this
sample is. Biases may occur in the data, in the schema, or during reasoning [Janowicz et al., 2018].
Examples of data biases include geographic biases that under-represent entities/relations from cer‐
tain parts of the world [Janowicz et al., 2018], linguistic biases that under-represent multilingual re‐
sources (e.g., labels and descriptions) for certain languages [Kaffee et al., 2017], social biases that
under-represent people of particular genders or races [Wagner et al., 2016], and so forth. In contrast,
schema biases may result from high-level definitions extracted from biased data [Janowicz et al.,
2018], semantic definitions that do not cover uncommon cases, etc. Unrecognised biases may lead to
adverse effects; for example, if the knowledge graph of Figure 7.1 has a geographic bias towards
events and attractions close to Santiago city – due perhaps to the sources used for creation, the em‐
ployment of curators from the city, etc. – then this may lead to tourism in and around Santiago being
disproportionately promoted to the detriment of tourism elsewhere in Chile. Measures of representa‐
tiveness may involve comparing known statistical distributions with those of the knowledge graph,
for example, comparing geolocated entities with known population densities [Janowicz et al., 2018],
linguistic distributions with known distributions of speakers [Kaffee et al., 2017], etc. Another more
general option is to compare the knowledge graph with general statistical laws, where Soulet et al.
[2018] use (non-)conformance with Benford’s law31 to measure representativeness in knowledge
note 31 Benford’s law states that the leading sign
graphs. bers is more likely to be small.
7.3 Coherency
Coherency refers to how well the knowledge graph conforms to – or is coherent with – the formal se‐
mantics and constraints defined at the schema-level.
Give us feedback!
https://kgbook.org 85/150
2/15/23, 5:13 PM Knowledge Graphs
7.3.1 Consistency
Consistency means that a knowledge graph is free of contradictions (i.e., inconsistencies) with respect
to the particular logical entailment considered. For example, if we apply the entailments defined in
Table 4.1 over the graph of Figure 7.1, we see that the edge Santiago de Chile –same as ➛ Santiago de Cuba
is inferred from both entities being the same as Santiago , which generates an inconsistency with the
edge Santiago de Chile –diff. from ➛ Santiago de Cuba as stated in the graph. While in this case it is evident
that Santiago de Cuba –same as ➛ Santiago is semantically inaccurate (considering that the venues con‐
nected to Santiago are in Chile), in other cases there may not be an obvious inaccuracy. Take, for ex‐
ample, the ontology defined in Figure 7.2, combined with the graph of Figure 7.1, and the ontologi‐
cal entailments of Tables 4.1–4.3. Noting that the food festival EID15 offers a takeaway service, ac‐
cording to the ontology, this entails that EID15 is a restaurant, a building, and a place, which is dis‐
joint with event. However, EID15 is also entailed to be a festival, and then an event, generating an
inconsistency. In this case there is no clear individual “error” leading to an inconsistency. Possibly
Table of Contents
the graph of Figure 7.1 should not use the property service for a food event (though it seems a “good
fit”), or perhaps the ontology of Figure 7.2 should not define the domain of the property service to be
a restaurant. Any ontological features in Tables 4.1–4.3 with a “not” condition can give rise to incon‐
sistencies if the negated condition is entailed. A measure of consistency can be the number of incon‐
sistencies found in a knowledge graph, possibly sub-divided into the number of such inconsistencies
identified by each semantic feature [Bonatti et al., 2011].
7.3.2 Validity
Validity means that the knowledge graph is free of constraint violations, such as captured by shape
expressions [Thornton et al., 2019] (see Section 3.1.2. We may, for example, specify a shape City
whose target nodes have at most one country. Then, taking the edges Chile ➛ country – Santiago –country
➛ Cuba from Figure 7.1, and assuming that Santiago becomes a target of City , we have a constraint
violation. Conversely, even if we defined analogous cardinality restrictions in an ontology (e.g., even
if we defined that country was functional), this would not necessarily cause an inconsistency since,
without UNA, we would first infer that Chile and Cuba refer to the same entity. Similarly, using
shapes, we can more easily detect missing data; for example, we can define a shape Event , and re‐
quire that it have at least one value for the property fare. Now, if EID15 becomes targetted by Event ,
then we will have a constraint violation as the node has no value for fare. Conversely, even if we de‐
fined analogous cardinality restrictions in an ontology (e.g., we defined that events have a minimum
cardinality of 1 for fare), this would not cause an inconsistency since, under the OWA, we would
rather entail that the event EID15 has some fair (that is not described in the graph). Consistency and
validity can thus indicate different types of issues. A straightforward measure of validity is to count
the number of violations per constraint.
7.4 Succinctness
Succinctness refers to the inclusion only of relevant content (avoiding “information overload”) that is
represented in a concise and intelligible manner.
7.4.1 Conciseness
Conciseness refers to avoiding schema and data elements that are irrelevant to the domain. Mendes et
al. [2012b] distinguish intensional conciseness (schema level), which refers to the case when the data
do not contain redundant schema elements (properties, classes, shapes, etc.), and extensional concise‐
ness (data level), where the data do not describe redundant entities and relations. For example, the
inclusion of a property and class for modelling jurisdictions and legal entities in the ontology of
Figure 7.2 may affect the intensional conciseness of the ontology in the context of a knowledge
Give
graph about tourist events. Similarly, the inclusion of data about Santiago de Cuba in our knowledge us feedback!
https://kgbook.org 86/150
2/15/23, 5:13 PM Knowledge Graphs
graph dedicated to tourism in Chile may affect the extensional conciseness of the knowledge graph,
potentially returning irrelevant results for the given domain. In general, conciseness can be measured
in terms of the ratio of properties, classes, shapes, entities, relations, etc., of relevance to the domain,
which may in turn require a gold standard, or measures of domain-relevance.
Representational conciseness refers to the extent to which content is compactly represented in the
knowledge graph, which may again be intensional or extensional [Zaveri et al., 2016]. For example,
having two properties category and type serving the same purpose would negatively affect the inten‐
sional form of representational conciseness, while having two nodes Santiago and Santiago de Chile
that split the data available about the capital of Chile would affect the extensional form of represen‐
tational conciseness. Another example of poor representational conciseness is the unnecessary use of
complex modelling constructs, such as using reification unnecessarily, or using linked lists when the
order of elements is not important [Hogan et al., 2012a]. An example of this is the anonymous node
used in Figure 7.1 to represent the days on which EID17 starts and ends, which could rather be di‐
Table of Contents
rectly associated with the event (at least if we assume that events have one start and one end mo‐
ment in time). A different example is the specification of the duration of EID15 , which could be cal‐
culated from the start and end values (assuming the correct datatypes were used). Though represen‐
tational conciseness is challenging to assess, measures such as the number of redundant nodes can be
used [Fürber and Hepp, 2011].
7.4.3 Understandability
Understandability refers to the ease with which data can be interpreted without ambiguity by human
users, which involves – at least – the provision of human-readable labels and descriptions (preferably
in different languages [Kaffee et al., 2017]) that allow such beings to understand what is being spo‐
ken about [Hogan et al., 2012a]. Referring back to Figure 7.1, though the nodes EID15 and EID17
are used to ensure unique identifiers for events, they should also be associated with labels, such as
Ñam . Ideally the human readable information is sufficient to disambiguate a particular node, such as
associating a description "Santiago, the capital of Chile"@en with Santiago to disambiguate the city from
synonymous ones. Measures of understandability may include the ratio of nodes with human-read‐
able labels and descriptions, the uniqueness of such labels and descriptions, the languages supported,
etc.
The list of quality dimensions provided here should be considered illustrative rather than complete.
Further dimensions may be pertinent in the context of specific domains, applications, or graph data
models. For more discussion, we refer to the survey by Zaveri et al. [2016] and to the book by Batini
and Scannapieco [2016].
Chapter 8
Refinement
Beyond assessing the quality of a knowledge graph, there exist techniques to refine the knowledge
graph, in particular to (semi-)automatically complete and correct the knowledge graph [Paulheim,
2017], aka knowledge graph completion and knowledge graph correction, respectively. As distinguished
from the creation and enrichment tasks outlined in Chapter 6, refinement typically does not involve
Give us feedback!
https://kgbook.org 87/150
2/15/23, 5:13 PM Knowledge Graphs
applying extraction or mappings over external sources in order to ingest their content into a given
knowledge graph (potentially using external sources to verify its content).
8.1 Completion
Knowledge graphs are characterised by incompleteness [West et al., 2014]. As such, knowledge
graph completion aims at filling in the missing edges (aka missing links) of a knowledge graph, i.e.,
edges that are deemed correct but are neither given nor entailed by the knowledge graph. This task
is often addressed with link prediction techniques proposed in the area of Statistical Relational
Learning [Getoor and Taskar, 2007], which predict the existence – or sometimes more generally, pre‐
dict the probability of correctness – of missing edges. For instance, one might predict that the edge
Moon Valley –bus ➛ San Pedro is a probable missing edge for the graph of Figure 5.2, given that most
bus routes observed are return services (i.e., bus is typically symmetric). Link prediction may target
three settings: general links involving edges with arbitrary labels, e.g., bus, flight, type, etc.; type links
involving edges with label type, indicating the type of an entity; and identity links involving edges
Table of Contents
with label same as, indicating that two nodes refer to the same entity (cf. Section 3.2.2). While type
and identity links can be addressed using general link prediction techniques, the particular semantics
of type and identity links can be addressed with custom techniques. (The related task of generating
links across knowledge graphs – referred to as link discovery [Nentwig et al., 2017] – will be dis‐
cussed later in Section 9.1.)
Link prediction, in the general case, is often addressed with inductive techniques as discussed in
Chapter 5, and in particular, knowledge graph embeddings and rule/axiom mining. For example,
given Figure 5.2, using knowledge graph embeddings, we may detect that given an edge of the form
x –bus ➛ y , a (missing) edge y –bus ➛ x has high plausibility, while using symbol-based approach‐
es, we may learn the high-level rule ?x –bus ➛ ?y ⇒ ?y –bus ➛ ?x that may infer/predict new bus
links. Either approach would help us to predict the missing link Moon Valley –bus ➛ San Pedro .
Type links are of particular importance to a knowledge graph, where dedicated techniques can be
leveraged taking into account the specific semantics of such links. In the case of type prediction,
there is only one edge label (type) and typically fewer distinct values (classes) than in other cases,
such that the task can be reduced to a traditional classification task [Paulheim, 2017], training mod‐
els to identify each semantic class based on features such as outgoing and/or incoming edge labels
on their instances in the knowledge graph [Paulheim and Bizer, 2013, Sleeman and Finin, 2013]. For
example, assume that in Figure 5.2 we also know that Arica , Calama , Puerto Montt , Punta Arenas and
Santiago are of type City . We may then predict that Iquique and Easter Island are also of type City
based on the presence of edges labelled flight to/from these nodes, which (we assume) are learnt to
be a good feature for prediction of that class (the former prediction is correct, while the latter is in‐
correct). Graph neural networks (see Section 5.3) can also be used for node classification/type
prediction.
Predicting identity links involves searching for nodes that refer to the same entity, but are not stated
or entailed to be the same; this is analogous to the task of entity matching (aka record linkage, dedu‐
plication, etc.) considered in more general data integration settings [Köpcke and Rahm, 2010]. Such
techniques are generally based on two types of matchers: value matchers determine how similar the
values of two entities on a given property are, which may involve similarity metrics on strings, num‐
bers, dates, etc.; while context matchers consider the similarity of entities based on various nodes and
edges [Köpcke and Rahm, 2010]. An illustrative example is given in Figure 8.1, where value match‐
ers will compute similarity between values such as 7400 and 7500 , while context matchers will com‐
Give us feedback!
https://kgbook.org 88/150
2/15/23, 5:13 PM Knowledge Graphs
pute similarity between Easter Island and Rapa Nui based on their surrounding information, such as
similar latitudes, longitudes, populations, and the same seat (conversely, a value matcher on this pair
of nodes would measure string similarity between “Easter Island” and “Rapa Ñui”).
Figure 8.1: Identity linking example: Easter Island and Rapa Nui denote the same place
A major challenge in this setting is efficiency, where a pairwise matching would require O(n2 ) com‐
parisons for n the number of nodes. To address this issue, blocking can be used to group similar enti‐
ties into (possibly overlapping, possibly disjoint) “blocks” based on similarity-preserving keys, with
matching performed within each block [Isele et al., 2011, Köpcke and Rahm, 2010, Draisbach and
Table of Contents
Naumann, 2011]; for example, if matching places based on latitude/longitude, blocks may represent
geographic regions. An alternative to discrete blocking is to use windowing over entities in a similari‐
ty-preserving ordering [Draisbach and Naumann, 2011], or to consider searching for similar entities
within multi-dimensional spaces (e.g., spacetime [Santipantakis et al., 2019], spaces with Minkowski
distances [Ngonga Ngomo, 2012], orthodromic spaces [Ngonga Ngomo, 2013], etc. [Sherif and
Ngonga Ngomo, 2018]). The results can either be pairs of nodes with a computed confidence of them
referring to the same entity, or crisp identity links extracted based on a fixed threshold, or binary
classification [Köpcke and Rahm, 2010]. For confident identity links, the nodes’ edges may then be
consolidated [Hogan et al., 2012b]; for example, we may select Easter Island as the canonical node and
merge the edges of Rapa Nui onto it, enabling us to find, e.g., World Heritage Sites in the Pacific Ocean
from Figure 8.1 based on the (consolidated) sub-graph World Heritage Site ➛ named – Easter Island –ocean ➛
Pacific .
8.2 Correction
As opposed to completion – which finds new edges in a knowledge graph – correction identifies and
removes existing incorrect edges in the knowledge graph. We here divide the principal approaches
for knowledge graph correction into two main lines: fact validation, which assigns a plausibility score
to a given edge, typically in reference to external sources; and inconsistency repairs, which aim to re‐
solve inconsistencies found in the knowledge graph through ontological axioms.
The task of fact validation (aka fact checking) [Gerber et al., 2015, Syed et al., 2018, Yin et al., 2008,
Syed et al., 2019, Esteves et al., 2018, Shiralkar et al., 2017, Shi and Weninger, 2016, Socher et al.,
2013, Bordes et al., 2013] involves assigning plausibility or veracity scores to facts/edges, typically
between 0 and 1. An ideal fact-checking function assumes a hypothetical reference universe (an ideal
knowledge graph) and would return 1 for the fact Santa Lucía –city ➛ Santiago (being true) while re‐
turning 0 for Sotomayor –city ➛ Santiago (being false). There is a clear relation between fact validation
and link prediction – with both relying on assessing the plausibility of edges/facts/links – and indeed
the same numeric- and symbol-based techniques can be applied for both cases. However, fact valida‐
tion often considers online assessment of edges given as input, whereas link prediction is often an
offline task that generates novel candidate edges to be assessed from the knowledge graph.
Furthermore, works on fact validation are characterised by their consideration of external reference
sources, which may be unstructured sources [Gerber et al., 2015, Syed et al., 2018, Samadi et al.,
2016, Yin et al., 2008] or structured sources [Syed et al., 2019, Shiralkar et al., 2017, Shi and
Weninger, 2016, Socher et al., 2013, Bordes et al., 2013].
Give us feedback!
https://kgbook.org 89/150
2/15/23, 5:13 PM Knowledge Graphs
Approaches based on unstructured sources assume that they are given a verbalisation function – using,
for example, rule-based approaches [Ngonga Ngomo et al., 2013, Ell et al., 2014], encoder–decoder
architectures [Gardent et al., 2017], etc. – that is able to translate edges into natural language.
Thereafter, approaches for computing the plausibility of facts in natural language – called fact find‐
ers [Pasternack and Roth, 2010, Pasternack and Roth, 2011] – can be directly employed. Many fact
finding algorithms construct an n-partite (often bipartite) graph whose nodes are facts and sources,
where a source is connected to a fact if the source “evidences” the fact, i.e., if it contains a text snip‐
pet that matches – with sufficient confidence – the verbalisation of the input edge. Two mutually-de‐
pendent scores, namely the trustworthiness of sources and the plausibility of facts, are then calculat‐
ed based on this graph, where fact finders differ on how they compute these scores [Pasternack and
Roth, 2011]. Here we mention three scores proposed by Pasternack and Roth [2010]
Sums [Pasternack and Roth, 2010] adapts the classical HITS centrality algorithm [Kleinberg,
1999] by defining sources as hubs (with 0 authority score) and facts as authorities (with 0 hub
score).
Average Log [Pasternack and Roth, 2010] extends HITS with a normalisation factor that pre‐
Table of Contents
vents a single source from receiving a high trustworthiness score by evidencing many facts
(that may be false).
Investment [Pasternack and Roth, 2010] lets the scores of facts grow with a non-linear function
based on “investments” coming from the connected sources. The score a source receives from a
fact is based on the individual facts in this particular source compared to the other connected
sources.
Pasternack and Roth [2011] then show that these three algorithms can be generalised into a single
multi-layered graph-based framework within which (1) a source can support a fact with a weight ex‐
pressing uncertainty, (2) similar facts can support each other, and (3) sources can be grouped togeth‐
er leading to an implicit support between sources of the same group. Other approaches for fact
checking of knowledge graphs later extended this framework [Galland et al., 2010, Samadi et al.,
2016]. Alternative approaches based on machine learning classifiers have also emerged, where com‐
monly-used features include trust scores for information sources, co-occurrences of facts in sources,
and so forth [Gerber et al., 2015, Syed et al., 2018].
Approaches for fact validation based on structured data typically assume external knowledge graphs
as reference sources and are based on finding paths that support the edge being validated.
Unsupervised approaches search for undirected [Shiralkar et al., 2017, Ciampaglia et al., 2015] or
directed [Syed et al., 2019] paths up to a given threshold length that support the input edge. The re‐
latedness between input edges and paths is computed using a mutual information function, such as
normalised pointwise mutual information [Bouma, 2009]. Supervised approaches rather extract fea‐
tures for input edges from external knowledge graphs [Sun et al., 2011, Zhao et al., 2015, Lao and
Cohen, 2010] and train a classification model to label the edges as true or false. An important set of
features are metapaths, which encode sequences of predicates that correlate positively with the edge
label of the input edge. Amongst such works, PredPath [Shi and Weninger, 2016] automatically ex‐
tracts metapaths based on type information. Several approaches rather encode the reference nodes
and edges using graph embeddings (see Section 5.2), which are then used to estimate the plausibility
of the input edge being validated.
Ontologies can contain axioms – such as disjointness – that lead to inconsistencies. While such ax‐
ioms can be provided by experts, they can can also be derived through symbolic learning, as dis‐
cussed in Section 5.4. Such axioms can then be used to detect inconsistencies. With respect to cor‐
recting a knowledge graph, however, detecting inconsistencies is not enough: techniques are also re‐
quired to repair such inconsistencies, which itself is not a trivial task. In the simplest case, we may
have an instance of two disjoint classes, such as that Santiago is of type City and Airport , which are
stated or found to be disjoint. To repair the inconsistency, it would be preferable to remove only the
“incorrect” class, but which should we remove? This is not a trivial question, particularly if we con‐
sider that one edge can be involved in many inconsistencies, and one inconsistency can involve
Give us feedback!
https://kgbook.org 90/150
2/15/23, 5:13 PM Knowledge Graphs
many edges. The issue of computing repairs becomes more complex when entailment is considered,
where we not only need to remove the stated type, but also all of the ways in which it might be en‐
tailed; for example, removing the edge Santiago –type ➛ Airport is insufficient if we further have an
edge Arica –flight ➛ Santiago combined with an axiom flight –range ➛ Airport . Töpper et al. [2012] sug‐
gest potential repairs for such violations – remove a domain/range constraint, remove a disjointness
constraint, remove a type edge, or remove an edge with a domain/range constraint – where one is
chosen manually. In contrast, Bonatti et al. [2011] propose an automated method to repair inconsis‐
tencies based on minimal hitting sets [Reiter, 1987], where each set is a minimal explanation for an
inconsistency. The edges to remove are chosen based on scores of the trustworthiness of their sources
and how many minimal hitting sets they are either elements of or help to entail an element of, where
the knowledge graph is revised to avoid re-entailment of the removed edges. Rather than repairing
the data, another option is to evaluate queries under inconsistency-aware semantics, such as return‐
ing consistent answers valid under every possible repair [Lukasiewicz et al., 2013].
In comparison to the quality clusters discussed in Chapter 7, the refinement methods discussed here‐
in address particular aspects of the accuracy, coverage, and coherency dimensions. Beyond these,
one could conceive of further refinement methods to address further quality issues of knowledge
graphs, such as succinctness. In general, however, the refinement tasks of knowledge graph completion
and knowledge graph correction have received the majority of attention until now. For further details
on knowledge graph refinement, we refer to the survey by Paulheim [2017].
Chapter 9
Publication
While it may not always be desirable to publish knowledge graphs (for example, those that offer a
competitive advantage to a company [Noy et al., 2019], it may be desirable or even required to pub‐
lish other knowledge graphs, such as those produced by volunteers [Vrandečić and Krötzsch, 2014,
Mahdisoltani et al., 2015, Lehmann et al., 2015], by publicly-funded research [Callahan et al., 2013,
Groth et al., 2014, The UniProt Consortium, 2014], by governmental organisations [Hendler et al.,
2012, Shadbolt and O'Hara, 2013]. Publishing refers to making the knowledge graph (or part there‐
of) accessible to the public, often on the Web. Knowledge graphs published as open data are called
open knowledge graphs (discussed in Section 10.1).
In the following, we first discuss two sets of principles that have been proposed to guide the publica‐
tion of data on the Web. We next discuss access protocols by which the public can interact with the
content of a knowledge graph. Finally, we consider techniques to restrict the access or usage of
(parts of) a knowledge graph.
We now discuss two key sets of publishing principles: the FAIR Principles Wilkinson et al. [2016],
and the Linked Data Principles [Berners-Lee, 2006].
Give us feedback!
https://kgbook.org 91/150
2/15/23, 5:13 PM Knowledge Graphs
The FAIR Principles were originally proposed in the context of publishing scientific data [Wilkinson
et al., 2016] – particularly motivated by maximising the impact of publicly-funded research – but the
principles generally apply to other situations where data are to be published in a manner that facili‐
tates their re-use by external agents, with particular emphasis on machine-readability.
FAIR itself is an acronym for four foundational principles, each with particular goals [Wilkinson et
al., 2016], that may apply to data, metadata, or both – the latter being denoted
note 32(meta)data. 32
Metadata are We
data about data. The distin
now describe the FAIR principles (slightly rephrasing the original wording sciences, where
in some in astronomy,
cases for for example, d
metadata may include coordinates and time.
brevity [Wilkinson et al., 2016]).
Findability refers to the ease with which external agents who might benefit from the dataset can
initially locate the dataset. Four sub-goals should be met:
F1: (meta)data are assigned a globally unique and persistent identifier.
F2: data are described with rich metadata (see R1).
F3: metadata explicitly include the identifier of the data they describe.
F4: (meta)data are registered or indexed in a searchable resource.
Table of Contents
Accessibility refers to the ease with which external agents can access the dataset (after locating
it). Two goals are defined, the first with two sub-goals:
A1: (meta)data are retrievable by their identifier via a standard protocol.
A1.1: the protocol is open, free, and universally implementable.
A1.2: the protocol uses authentication and authorisation if suitable.
A2: metadata are accessible, even when the data are no longer available.
Interoperability refers to the ease with which the dataset can be exploited (in unison with other
datasets) using standard tools. Three goals are defined:
I1: meta)data use an accessible, agreed-upon, and general knowledge representation
formalism.
I2: (meta)data use vocabularies that follow FAIR principles.
I3: (meta)data include qualified references to other (meta)data.
Reusability refers to the ease with which the dataset can be re-used in conjunction with other
datasets. One goal is defined (with three sub-goals):
R1: meta(data) are richly described with accurate and relevant attributes.
R1.1. (meta)data are released with a clear and accessible license.
R1.2. (meta)data are associated with detailed provenance.
R1.3. (meta)data meet domain-relevant community standards.
In the context of knowledge graphs, a variety of vocabularies, tools, and services have been proposed
that both directly and indirectly help to satisfy the FAIR principles. In terms of Findability, as dis‐
cussed in Chapter 2, IRIs are built into the RDF model, providing a general schema for global identi‐
fiers. In addition, resources such as the Vocabulary of Interlinked Datasets (VoID) [Alexander et al.,
2009] allow for representing metadata about graphs, while services such as DataHub [Bhardwaj et
al., 2015] provide a central repository of such dataset descriptions. Access protocols that enable
Accessibility will be discussed in Section 9.2, while mechanisms for authorisation will be discussed in
Section 9.3. With respect to Interoperability, as discussed in Chapter 4, ontologies serve as a general
knowledge representation formalism, and can in turn be used to describe vocabularies that follow
FAIR principles. Regarding Reusability, licensing will be discussed in Section 9.3, while the PROV
Data Model [Gil et al., 2013] discussed in Chapter 3, can encode provenance in detail.
Various knowledge graphs have been published using FAIR principles, where Wilkinson et al. [2016]
explicitly mention Open PHACTS [Groth et al., 2014], a data integration platform for drug discovery,
and UniProt [The UniProt Consortium, 2014], a large collection of protein sequence and annotation
data, as conforming to FAIR principles. Both datasets offer graph views of their content through RDF.
Wilkinson et al. [2016] state that FAIR Principles “precede implementation choices”, meaning that
the principles do not cover how they can or should be achieved. Preceding the FAIR Principles by al‐
most a decade are the Linked Data Principles, proposed by Berners-Lee [2006], which provide a
Give us feedback!
https://kgbook.org 92/150
2/15/23, 5:13 PM Knowledge Graphs
technical basis for one possible way in which these FAIR Principles can be achieved. Specifically the
Linked Data Principles are as follows:
These principles were proposed in a Semantic Web setting, where for principle (3), the standards
based on RDF (including RDFS, OWL, etc.) are currently recommended for use, particularly because
they allow for naming entities using HTTP IRIs, which further paves the way for satisfying all four
principles. As such, these principles outline a way in which (RDF) graph-structured data can be pub‐
lished on the Web such that these graphs are interlinked to form what Berners-Lee [2006] calls a
“Web of Data”, whose goal is to increase automation on the Web by making content available not
only in (HTML) documents intended for human consumption, but also as (RDF) structured data that
machines can locate, retrieve, combine, validate, reason over, query over, etc., towards solving tasks
Table of Contents
automatically [Hogan, 2020b]. Conceptually, the Web of Data is then composed of graphs of data
published on individual web-pages, where one can click on a node or edge-label – or more precisely
perform a HTTP lookup on an IRI of the graph – to be transported to another graph elsewhere on the
Web with relevant content for that node or edge-label, and so on recursively.
Figure 9.1 provides a small example with two Linked Data documents published on the Web, with
each containing an RDF graph. As discussed in Section 3.2, terms such as clv:Concert,
wd:Q142701, rdfs:label, etc., are abbreviations for IRIs, where, for example, wd:Q142701 ex‐
pands to http://www.wikidata.org/entity/Q142701. Prefixes beginning with cl are fictitious prefixes we
assume to have been created by the Chilean tourist board. The IRIs prefixed with ↪ indicate the
document returned if the node is looked up. The leftmost document is published by the tourist board
and describes Lollapalooza 2018 (identified by the node cle:LP2018 ), which links to the headlining
act Pearl Jam ( wd:Q142701 ) described by an external knowledge graph, namely Wikidata. By looking
up the node wd:Q142701 in the leftmost graph, the IRI dereferences (i.e., returns via HTTP) the docu‐
ment with the RDF graph on the right describing that entity in more detail. From the rightmost doc‐
ument, the node wd:Q221535 can be looked up, in turn, to find a graph about Eddie Vedder (not
shown in the example). The IRIs for entities and documents are distinguished to ensure that we do
not confuse data about the entity and the document; for example, while
wd:Q221535 refers to Eddie
Vedder, the IRI wdd:Q221535 refers to the document about Eddie Vedder; if we were to assign a
last-modified date to the document, we should use wdd:Q221535 not wd:Q221535 . In Figure 9.1, we
can further observe that edge labels (which are also IRIs) and nodes representing classes (e.g.,
clv:Concert ) can also be dereferenced, typically returning semantic definitions of the respective terms.
Figure 9.1: Two example Linked Data documents from two websites, each containing an RDF graph, where
wd:Q142701 refers to Pearl Jam in Wikidata while wdd:Q142701 refers to the RDF graph about Pearl
Jam, and where wd:Q221535 refers to Eddie Vedder while wdd:Q221535 refers to the RDF graph about
Eddie Vedder; the edge-label wdt:571 refers to “inception” in Wikidata, while wdt:527 refers to “has part”
A key challenge is posed by the fourth principle – include links to related entities – as illustrated in
Figure 9.1, where wd:Q221535 in the leftmost graph constitutes a link to related content about Pearl
Jam in an external knowledge graph. Specifically, the link discovery task considers adding such links
Give us feedback!
https://kgbook.org 93/150
2/15/23, 5:13 PM Knowledge Graphs
from one knowledge graph to another, which may involve inclusion of IRIs that dereference to exter‐
nal graphs (per Figure 9.1), or links with special semantics such as identity links. In comparison with
the link prediction task discussed in Section 8.1, which is used to complete links within a knowledge
graph, link discovery aims to discover links across knowledge graphs, which involves unique aspects:
first, link discovery typically considers disjoint sets of source (local) nodes and target (remote)
nodes; second, the knowledge graphs may often use different vocabularies; third, while in link pre‐
diction there already exist local examples of the links to predict, in link discovery, there are often no
existing links between knowledge graphs to learn from. A common technique is to define manually-
crafted linkage rules (aka link specifications) that apply heuristics for defining links that potentially
incorporate similarity measures [Ngonga Ngomo and Auer, 2011, Volz et al., 2009]. Link discovery
is greatly expedited by the provision of standard identifier schemes within knowledge graphs, such
as ISBNs for books, alpha-2 and alpha-3 codes for countries (e.g., cl, clp), or even links to common
knowledge graphs such as DBpedia [Lehmann et al., 2015] or Wikidata [Vrandečić and Krötzsch,
2014] (that themselves include standard identifiers). We refer to the survey on link discovery by
Nentwig et al. [2017] for more details.
Table of Contents
More finer-grained recommendations for publishing Linked Data have also been proposed, relating to
how best to implement dereferencing, what kinds of links to include, how to publish and interlink
vocabularies, amongst other considerations [Heath and Bizer, 2011, Janowicz et al., 2014]. We refer
to the book by Heath and Bizer [2011] for more discussion on how to publish Linked Data on the
Web.
Publishing involves giving access to the public to interact with the knowledge graph, which implies
the provision of access protocols that define the requests that agents can make and the response that
they can expect as a result. Per the Accessibility principle of FAIR (specifically A1.1), this protocol
should be open, free, and universally implementable. In the context of knowledge graphs, as shown
in Figure 9.2, there are a number of access protocols to choose from, varying from simple protocols
that allow users to simply download all content, towards protocols that accept and evaluate increas‐
ingly complex requests. While simpler protocols require less computation on the server that publish‐
es the data, more complex protocols allow agents to request more specific data, thus reducing band‐
width. A knowledge graph may also offer a variety of access protocols catering to different agents
with different requirements [Verborgh et al., 2014]. We now discuss such access protocols.
Figure 9.2: Access protocols for knowledge graphs, from simple protocols (left) to more complex protocols
(right)
9.2.1 Dumps
A dump is a file or collection of files containing the content of the knowledge graph available for
download. The request in this case is for the file(s) and the response is the content of the file(s). In
order to publish dumps, first of all, concrete – and ideally standard – syntaxes are required to seri‐
alise the graph. While for RDF graphs there are various standard syntaxes available based on
XML [Gandon and Schreiber, 2014], JSON [Sporny et al., 2014], custom syntaxes [Prud'hommeaux
and Carothers, 2014], and more besides, currently there are only non-standard syntaxes available for
property graphs [Tomaszuk et al., 2019]. Second, to reduce bandwidth, compression methods can be
applied. While standard compression such as GZIP or BZip2 can be straightforwardly applied on any
file, custom compression methods have been proposed for graphs that not only offer better compres‐
Give
sion ratios than these standard methods, but also offer additional functionalities, such as compact in‐ us feedback!
https://kgbook.org 94/150
2/15/23, 5:13 PM Knowledge Graphs
dexes for performing efficient lookups once the file is downloaded [Fernández et al., 2013]. Finally,
to further reduce bandwidth, when the knowledge graph is updated, “diffs” can be computed and
published to obviate the need for agents to download all data from scratch (see [Tummarello et al.,
2007, Papavasileiou et al., 2013, Ahn et al., 2015]). Still, however, dumps are only suited to certain
use-cases, in particular for agents that wish to maintain a full local copy of a knowledge graph. If an
agent were rather only interested in, for example, all food festivals in Santiago, downloading the en‐
tire dump may require transferring and processing a lot of irrelevant data.
Protocols for performing node lookups accept a node (id) request (e.g., cle:LP2018 in Figure 9.1) and
return a (sub-)graph describing that node (e.g., the document cld:LP2018). Such a protocol is the
basis for the Linked Data principles outlined previously, whereby node lookups are implemented
through HTTP dereferencing, which further allows nodes in remote graphs to be referenced from
across the Web. Although there are varying definitions on what content should be returned for a
node [Stickler, 2005], a common convention is to return a sub-graph containing either all outgoing
Table of Contents
edges for that node or all incident edges (both outgoing and incoming) for that node [Hogan et al.,
2012a]. Though simple, mechanisms for evaluating graph patterns can be implemented on top of a
node lookup interface by traversing from node to node per the particular graph pattern [Hartig et al.,
2009]; for example, to find all food festivals in Santiago – represented by the graph pattern
Food Festival ➛ type – ?ff –location ➛ Santiago – we may perform a node lookup for Santiago , subsequently
performing a node lookup for each node connected by a location edge to Santiago , returning those
nodes declared to be of type Food Festival . However, such an approach may not be feasible if no start‐
ing node is declared (e.g., if all nodes are variables), if the node lookup service does not return in‐
coming edges, etc. The client agent may also need to request more data than necessary; for example,
the document returned for Santiago may return a lot of data irrelevant to the query, and nodes with
a location in Santiago that are not instances of Food Festival still need to be looked up to check their
type. Node lookups are relatively inexpensive for servers to support in terms of CPU, but may again
waste bandwidth due to transferring irrelevant data.
Edge patterns – also known as triple patterns in the case of directed, edge-labelled graphs – are single‐
ton graph patterns, i.e., graph patterns with a single edge. Examples of edge patterns are ?ff –type ➛
Food Festival or ?ff –location ➛ Santiago , etc., where any term can be a variable or a constant. A proto‐
col for edge patterns accepts such a pattern and returns all solutions for the pattern. Edge patterns
provide more flexibility than node lookups, where graph patterns are more readily decomposed into
edge patterns than node lookups. With respect to the agent interested in food festivals in Santiago,
they can first, for example, request solutions for the edge pattern ?ff –location ➛ Santiago and locally
join/intersect these solutions with those of ?ff –type ➛ Food Festival . Given that some edge patterns
(e.g., ?x –?y➛ ?z ) can return many solutions, protocols for edge patterns may offer additional practi‐
cal features such as iteration or pagination over results [Verborgh et al., 2016]. Much like node
lookups, the server cost of responding to a request is relatively low and easy to predict. However, the
server may often need to transfer irrelevant intermediate results to the client, which in the previous
example may involve returning nodes located in Santiago that are not food festivals. This issue is fur‐
ther aggravated if the client does not have access to statistics about the knowledge graph in order to
plan how to best perform the join; for example, if there are relatively few food festivals but many
things located in Santiago, rather than intersecting the solutions of the two aforementioned edge pat‐
terns, it should be more efficient to send a request for each food festival to see if it is in Santiago, but
deciding this requires statistics about the knowledge graph. Extensions to the edge-pattern protocol
have thus been proposed to allow for more efficient joins [Hartig et al., 2017], such as allowing
batches of solutions to be sent alongside the edge pattern to only return solutions compatible with
the solutions in the request [Hartig and Buil Aranda, 2016] (e.g., sending a batch of solutions for ?ff
–type ➛ Food Festival to join with the solutions for the request ?ff –location ➛ Santiago ).
https://kgbook.org 95/150
2/15/23, 5:13 PM Knowledge Graphs
Another alternative is to let client agents make requests based on (complex) graph patterns (see
Section 2.2), with the server returning (only) the final solutions. In our running example, this in‐
volves the client issuing a request for Food Festival ➛ type – ?ff –location ➛ Santiago and directly receiving
the relevant results. Compared with the previous protocols, this protocol is much more efficient in
terms of bandwidth: it allows clients to make more specific requests and the server to return more
specific responses. However, this reduction in bandwidth use comes at the cost of the server having
to evaluate much more complex requests, where, furthermore, the costs of a single request are much
more difficult to anticipate. While a variety of optimised engines exist for evaluating (complex)
graph patterns (e.g., [Erling, 2012, Miller, 2013, Thompson et al., 2014] amongst many others), the
problem of evaluating such queries is known to be intractable [Angles et al., 2017]. Perhaps for this
reason, public services offering such a protocol (most often supporting SPARQL queries [Harris et al.,
2013]) have been found to often exhibit downtimes, timeouts, partial results, slow performance,
etc. [Buil-Aranda et al., 2013b]. Even considering such issues, however, popular services continue to
receive – and successfully evaluate – millions of requests/queries per day [Malyshev et al., 2018,
Saleem et al., 2015], with difficult (worst-case) instances being rare in practice [Bonifati et al.,
2017].
Table of Contents
While Figure 9.2 makes explicit reference to some of the most commonly-encountered access proto‐
cols found for knowledge graphs in practice, one may of course imagine other protocols lying almost
anywhere on the spectrum from more simple to more complex interfaces. To the right of (Complex)
Graph Patterns, one could consider supporting even more complex requests, such as queries with en‐
tailments [Glimm, 2011], queries that allow recursion [Reutter et al., 2015], federated queries that
can join results from remote services [Buil-Aranda et al., 2013a], or even (hypothetically) supporting
Turing-complete requests that allow running arbitrary procedural code on a knowledge graph. As
mentioned at the outset, a server may also choose to support multiple, complementary
protocols [Verborgh et al., 2014].
Considering our hypothetical tourism knowledge graph, at first glance, one might assume that the
knowledge required to deliver the envisaged services is public and thus can be used both by the
tourism board and the tourists. On closer inspection, however, we may see the need for usage control
in various forms:
both the tourist board and its partners should associate an appropriate license with knowledge
that they contribute to the knowledge graph, such that the terms of use are clear to all interest‐
ed parties;
a tourist might opt to install an app on their mobile phone that could be used to recommend
tourist attractions based on their location, bringing with it potential privacy concerns regarding
who has access to their location;
the tourist board may be required to report criminal activities to the police services and thus
may need to encrypt personal information; and
the tourist board could potentially share information relating to tourism demographics in an
anonymous format to allow for other agencies and companies to anticipate demand and im‐
prove transport infrastructure on strategic routes.
Thus in this section, we examine the state of the art in terms of knowledge graph licensing, usage
policies, encryption, and anonymisation.
9.3.1 Licensing
When it comes to associating machine readable licenses with knowledge graphs, the W3C Open
Digital Rights Language (ODRL) [Iannella and Villata, 2018] provides an information model and re‐
Give
lated vocabularies that can be used to specify permissions, duties, and prohibitions with respect to us feedback!
https://kgbook.org 96/150
2/15/23, 5:13 PM Knowledge Graphs
actions relating to assets. ODRL supports fine-grained descriptions of digital rights that are represent‐
ed as – and thus can be embedded within – graphs. Figure 9.3 illustrates a license granting the as‐
signee the permission to Modify , Distribute , and Derive work from the Event Graph (e.g., Figure 2.1);
however the assignee is obliged to Attribute the copyright holder. From a modelling perspective,
ODRL can be used to model several well-known license families, for instance Apache, Creative
Commons (CC), and Berkeley Software Distribution (BSD), to name but a few [Cabrio et al., 2014,
Panasiuk et al., 2018]. Additionally, Cabrio et al. [2014] propose methods to automatically extract
machine-readable licenses from unstructured text. From a reasoning perspective, license compatibili‐
ty validation and composition techniques [Villata and Gandon, 2012, Governatori et al., 2013,
Moreau et al., 2019] can be used to combine knowledge graphs that are governed by different licens‐
es. Such techniques are employed by the the Data Licenses Clearance Center (DALICC), which in‐
cludes a library of standard machine readable licenses, and tools that enable users both to compose
arbitrary custom licenses and also to verify the compatibility of different licenses [Pellegrini et al.,
2019].
Table of Contents
Figure 9.3: A license for event data, along with permissions, actions, and obligations
Access control policies based on edge patterns can be used to restrict access to parts of a knowledge
graph [Reddivari et al., 2005, Flouris et al., 2010, Kirrane et al., 2013]. WebAccessControl (WAC)33
is an access control framework for graphs that uses WebID for authentication and provides a vocabu‐
lary for specifying access control policies. Extensions of this WAC vocabulary have been proposed to
note 33 WAC,
capture privacy preferences [Sacco and Passant, 2011] and to cater for contextual
http://www.w3.org/wiki/WebAccessC
constraints [Villata et al., 2011, Costabello et al., 2012]. Although ODRL is primarily used to specify
licenses, profiles to additionally specify access policies [Steyskal and Polleres, 2014] and regulatory
obligations [Agarwal et al., 2018, De Vos et al., 2019] have also been proposed in recent years, as
discussed in the survey by Kirrane et al. [2017].
As a generalisation of access policies, usage policies specify how data can be used: what kinds of pro‐
cessing can be applied, by whom, for what purpose, etc. The example usage policy presented in
Figure 9.4 states that the process Analyse of Location Graph can be performed on Internal Servers by
members of Company Staff in order to provide Event Recommendations . Vocabularies for usage policies
have been proposed by the SPECIAL H2020 project [Bonatti et al., 2019] and the W3C Data Privacy
Vocabularies and Controls Community Group (DPVCG) [Pandit et al., 2019, Bonatti and Kirrane,
2019]. Once specified in these vocabularies, usage policies can then be used to verify that data pro‐
cessing conforms to legal norms and to the consent provided by subjects [Delanaux et al., 2018,
Bonatti and Kirrane, 2019].
Figure 9.4: A policy for usage of a sub-graph of location data in the knowledge graph
https://kgbook.org 97/150
2/15/23, 5:13 PM Knowledge Graphs
Rather than internally controlling usage, the tourist board could use encryption mechanisms on parts
of the published knowledge graph, for example relating to reports of crimes, and provide keys to
partners who should have access to the plaintext. While a straightforward approach is to encrypt the
entire graph (or sub-graphs) with one key, more fine-grained encryption can be performed for indi‐
vidual nodes or edge-labels in a graph, potentially providing different clients access to different in‐
formation through different keys [Giereth, 2005]. The CryptOntology [Gerbracht, 2008] can further
be used to embed details about the encryption mechanism used within the knowledge graph.
Figure 9.5 illustrates how this could be used to encrypt the names of claimants from Figure 6.4, stor‐
ing the ciphertext zhk…kjg , as well as the key-length and encryption algorithm used. In order to
grant access to the plaintext, one approach is to encrypt individual edges with symmetric keys so as
to allow specific types of edge patterns to only be executed by clients with the appropriate
key [Kasten et al., 2013]. This approach can be used, for example, to allow clients who know a
claimant ID (e.g., Claimant-XY12SDA ) and have the appropriate key to find (only) the name of the
claimant through an edge pattern Claimant-XY12SDA –Claimant-name ➛ ?name . A key limitation of this
approach, however, is that it requires attempting to decrypt all edges to find all possible solutions. A
more efficient alternative is to combine functional encryption and specialised indexing to retrieve so‐
Table of Contents
lutions from the encrypted graph without attempting to decrypt all edges [Fernández et al., 2017].
Figure 9.5: Directed edge-labelled graph with the name of the claimant encrypted; plaintext elements are
dashed and may be omitted from published data (possibly along with encryption details)
9.3.4 Anonymisation
Consider that the tourist board acquires information on transport taken by individuals within the
country, which can be used – not only by the board, but potentially other stakeholders, such as trav‐
el companies – to understand trajectories taken by tourists. However, from a data-protection perspec‐
tive, it would be advisable to anonymise the knowledge graph to avoid leaking the personal travel
history of individuals.
A first approach to anonymisation is to suppress and generalise knowledge in a graph such that indi‐
viduals cannot be identified, based on k -anonymity [Samarati and Sweeney, note 34 34
1998] , l-diversity guarantees
k -anonymity [Li that the data of
35
et al., 2007] , etc. Approaches that apply k -anonymity on graphs identify andleast k − 1 other
suppress
note 35 individuals.
“quasi-iden‐
l-diversity guarantees that sensitive data
tifiers” that would allow a given individual to be distinguished from fewer than
each
k− group
1 otherof individuals;
individu‐ this avoids leaks such
of individuals)
als [Radulovic et al., 2015, Heitmann et al., 2017]. Figure 9.6 illustrates a possible result in the
of kdata
- have been pick-pock
reveal sensitive information about individuals fr
anonymisation for a sub-graph describing a flight passenger, where quasi-identifiers (passport, plane
ticket) have been converted into blank nodes, ensuring that the passenger (the dashed blank node)
cannot be distinguished from k − 1 other individuals. In the context of a graph, however, neighbour‐
hood attacks [Zhou and Pei, 2011] – using information about neighbours – can also break k -anonymi‐
ty, where we also suppress the day and time of the flight, which, though not sensitive information
per se, could otherwise break k -anonymity for passengers (if, for example, a particular flight had
fewer than k males from the U.S. onboard). The graph shown in Figure 9.6 then offers k -anonymity
for the particular individual assuming that at least k male passengers from the U.S. flew during
December 2018 from Arica to Santiago.
Give us feedback!
https://kgbook.org 98/150
2/15/23, 5:13 PM Knowledge Graphs
Figure 9.6: Anonymised sample of a directed edge-labelled graph describing a passenger (dashed) of a
flight
More complex neighbourhood attacks may rely on more abstract graph patterns, observing that indi‐
viduals can be deanonymised purely from knowledge of the graph structure, even if all nodes and
edge labels are left blank; for example, if we know that a team of k − 1 players take flights together
for a particular number of away games, we could use this information for a neighbourhood attack
that reveals the set of players in the graph. Hence a number of guarantees specific to graphs have
been proposed, including k -degree anonymity [Liu and Terzi, 2008], which ensures that individuals
Table of Contents
cannot be deanonymised by attackers with knowledge of the degree of particular individuals. The
approach is based on minimally modifying the graph to ensure that each node has at least k − 1 oth‐
er nodes with the same degree. A stronger guarantee, called k -isomorphic neighbour
anonymity [Zhou and Pei, 2008], avoids neighbourhood attacks where an attacker knows how an in‐
dividual is connected to nodes in their neighbourhood; this is done by modifying the graph to ensure
that for each node, there exist at least k − 1 nodes with isomorphic (i.e., identically structured)
neighbourhoods elsewhere in the graph. Both approaches only protect against attackers with knowl‐
edge of bounded neighbourhoods. An even stronger notion is that of k -automorphism [Zou et al.,
2009], which ensures that for every node, it is structurally indistinguishable from k − 1 other nodes,
thus avoiding any attack based on structural information (as a trivial example, a k -clique or a k -cycle
satisfy k -automorphism). Many of these techniques for anonymisation of graph data were motivated
by social networks [Narayanan and Shmatikov, 2009], though they can also be applied to knowledge
graphs, per the work of Lin and Tripunitara [2017], who adapt k -automorphism for directed edge-la‐
belled graphs (specifically RDF graphs).
While the aforementioned approaches anonymise data, a second approach is to apply anonymisation
when answering queries, such as adding noise to the solutions in a way that preserves privacy. One
approach is to apply ε-differential privacy [Dwork, 2006]36 for querying graphs
note [Silva et al., 2017].
36 ε-differential privacy ensures that the pr
Such mechanisms are typically used for aggregate (e.g., count) queries, where (e.g., noise
query)isapplied
added to data, to which random
when
avoid leaks about individuals. To illustrate, differential privacy may allow for the data
counting theincludes
numberor excludes any individu
of passengers of specified nationalities taking specified flights, adding (just enough) random noise to
the count to ensure that we cannot tell, within a certain probability (controlled by ε), whether or not
a particular individual took a flight, where, intuitively speaking, we would require (proportionally)
less noise for nationalities with many passengers in the data, but more noise to “hide” passengers
from more uncommon nationalities.
These approaches require information loss for stronger guarantees of privacy; which to choose is
thus heavily application dependent. If the anonymised data are to be published in their entirety as a
“dump”, then an approach based on k -anonymity can be used to protect individuals, while l-diversity
can be used to protect groups. On the other hand, if the data are to be made available, in part,
through a query interface, then ε-differential privacy is a more suitable framework.
Chapter 10
https://kgbook.org 99/150
2/15/23, 5:13 PM Knowledge Graphs
In this chapter, we discuss some of the most prominent knowledge graphs that have emerged in the
past years. We begin by discussing open knowledge graphs, most of which have been published on
the Web per the guidelines and protocols described in Chapter 9. We later discuss enterprise knowl‐
edge graphs that have been created by companies from diverse industries for a wide range of
applications.
By open knowledge graphs, we refer to knowledge graphs published under the Open Data philosophy,
namely that “open means anyone can freely access, use, modify, and share for any purpose (subject, at
most, to requirements that preserve provenance and openness)”.37 Many open knowledge
note 37 See graphs have
http://opendefinition.org/
been published in the form of Linked Open Datasets [Heath and Bizer, 2011], which are (RDF) graphs
published under the Linked Data principles (see Section 9.1.2) following the Open Data philosophy.
Many of the most prominent open knowledge graphs – including DBpedia [Lehmann et al., 2015],
YAGO [Suchanek et al., 2007], Freebase [Bollacker et al., 2007b], and Wikidata [Vrandečić and
Table of Contents
Krötzsch, 2014] – cover multiple domains, representing a broad diversity of entities and relation‐
ships; we first discuss these in turn. Later we discuss some of the other (specific) domains for which
open knowledge graphs are currently available. Most of the open knowledge graphs we discuss in
this section are modelled in RDF, published following Linked Data principles, and offer access to
their data through dumps (RDF), node lookups (Linked Data), graph patterns (SPARQL) and, in some
cases, edge patterns (Triple Pattern Fragments).
10.1.1 DBpedia
The DBpedia project was developed to extract a graph-structured representation of the semi-struc‐
tured data embedded in Wikipedia articles [Auer et al., 2007], enabling the integration, processing,
and querying of these data in a unified manner. The resulting knowledge graph is further enriched
by linking to external open resources, including images, webpages, and external datasets such as
DailyMed, DrugBank, GeoNames, MusicBrainz, New York Times, and WordNet [Lehmann et al.,
2015]. The DBpedia extraction framework consists of several components, corresponding to abstrac‐
tions of Wikipedia article sources, graph storage and serialisation destinations, wiki-markup extrac‐
tors, parsers, and extraction managers [Bizer et al., 2009]. Specific extractors are designed to process
labels, abstracts, interlanguage links, images, redirects, disambiguation pages, external links, internal
pagelinks, homepages, categories, and geocoordinates. The content in the DBpedia knowledge graph
is not only multidomain, but also multilingual: as of 2012, DBpedia contained labels and abstracts in
up to 97 different languages [Mendes et al., 2012a]. Entities within DBpedia are classified using four
different schemata in order to address varying requirements [Bizer et al., 2009]. These schemata in‐
clude a Simple Knowledge Organization System (SKOS) representation of Wikipedia categories, a Yet
Another Great Ontology (YAGO) classification schema (discussed presently), an Upper Mapping and
Binding Exchange Layer (UMBEL) ontology categorisation schema, and a custom schema called the
DBpedia ontology with classes such as Person, Place, Organisation, and Work [Lehmann et al.,
2015]. DBpedia also supports live synchronisation in order to remain consistent with dynamic
Wikipedia articles [Lehmann et al., 2015].
YAGO likewise extracts graph-structured data from Wikipedia, which are then unified with the hier‐
archical structure of WordNet to create a “light-weight and extensible ontology with high quality and cov‐
erage” [Suchanek et al., 2007]. This knowledge graph aims to be applied for various information
technology tasks, such as machine translation, word sense disambiguation, query expansion, docu‐
ment classification, data cleaning, information integration, etc. While earlier approaches automati‐
cally extracted structured knowledge from text using pattern matching, natural language processing
(NLP), and statistical learning, the resulting content tended to lack in quality when compared with
what was possible through manual construction [Suchanek et al., 2007]. However, manual construc‐
tion is costly, making it challenging to achieve broad coverage and keep the data up-to-date. In order
Give us feedback!
https://kgbook.org 100/150
2/15/23, 5:13 PM Knowledge Graphs
to extract data with high coverage and quality, YAGO (like DBpedia) mostly extracts data from
Wikipedia infoboxes and category pages, which contain core entity information and lists of articles
for a specific category, respectively. These, in turn, are unified with hierarchical concepts from
WordNet [Suchanek et al., 2008]. A schema – called the YAGO model – provides a vocabulary de‐
fined in RDFS; this model allows for representing words as entities, capturing synonymy and ambigu‐
ity [Suchanek et al., 2007]. The model further supports reification, n-ary relations, and data
types [Suchanek et al., 2008]. Refinement mechanisms employed within YAGO include canonicalisa‐
tion, where each edge and node is mapped to a unique identifier and duplicate elements are re‐
moved, and type checking, where nodes that cannot be assigned to a class by deductive or inductive
methods are eliminated [Suchanek et al., 2008]. YAGO would be extended in later years to support
spatio-temporal context [Hoffart et al., 2011] and multilingual Wikipedias [Mahdisoltani et al.,
2015].
10.1.3 Freebase
Freebase was a general-purpose, broad collection of human knowledge that aimed to address some of
Table of Contents
the large-scale information integration problems associated with the decentralised nature of the
Semantic Web, such as uneven adoption, implementation challenges, and distributed query perfor‐
mance limitations [Bollacker et al., 2007a]. Unlike DBpedia and YAGO – which are mostly extracted
from Wikipedia/WordNet – Freebase solicited contributions directly from human editors. Included in
the Freebase platform were a scalable data store with versioning mechanisms; a large data object
store (LOB) for the storage of text, image, and media files; an API that could be queried using the
Metaweb Query Language (MQL); a Web user interface; and a lightweight typing system [Bollacker
et al., 2007a]. The latter typing system was designed to support collaborative processes. Rather than
forcing ontological correctness or logical consistency, the system was implemented as a loose collec‐
tion of structuring mechanisms – based on datatypes, semantic classes, properties, schema defini‐
tions, etc. – that allowed for incompatible types and properties to coexist simultaneously [Bollacker
et al., 2007a]. Content could be added to Freebase interactively through the Web user interface or in
an automated way by leveraging the API’s write functionality. Freebase had been acquired by Google
in 2010, where the content of Freebase formed an important part of the Google Knowledge Graph
announced in 2012 [Singhal, 2012]. When Freebase became read-only as of March 2015, the knowl‐
edge graph contained over three billion edges. Much of this content was subsequently migrated to
Wikidata [Pellissier Tanon et al., 2016].
10.1.4 Wikidata
Wikipedia contains a wealth of semi-structured data embedded in info-boxes, lists, tables, etc., as ex‐
ploited by DBpedia and YAGO. However, these data have traditionally been curated and updated
manually across different articles and languages; for example, a goal scored by a Chilean football
player may require manual updates in the player's article, the tournament article, the team article,
lists of top scorers, and so forth, across hundreds of language versions. Manual curation has led to a
variety of data quality issues, including contradictory data in different articles, languages, etc. The
Wikimedia Foundation uses Wikidata as a centralised, collaboratively-edited knowledge graph to
supply Wikipedia – and arbitrary other clients – with data. Under this vision, a fact could be added
to Wikidata once, triggering the automatic update of potentially multitudinous articles in Wikipedia
across different languages [Vrandečić and Krötzsch, 2014]. Like Wikipedia, Wikidata is also consid‐
ered a secondary source containing claims that should reference primary sources, though claims can
also be initially added without reference [Piscopo et al., 2017]. Wikidata further allows for different
viewpoints in terms of potentially contradictory (referenced) claims [Vrandečić and Krötzsch, 2014].
Wikidata is multilingual, where nodes and edges are assigned language-agnostic Qxx and Pxx codes
(see Figure 9.1) and are subsequently associated with labels, aliases, and descriptions in various lan‐
guages [Kaffee et al., 2017], allowing claims to be surfaced in these languages. Collaborative editing
is not only permitted on the data level, but also on the schema level, allowing users to add or modify
lightweight semantic axioms [Piscopo and Simperl, 2018] – including sub-classes, sub-properties, in‐
verse properties, etc. – as well as shapes [Boneva et al., 2019]. Wikidata offers various access proto‐
cols [Malyshev et al., 2018] and has received broad adoption, being used by Wikipedia to generate
Give us feedback!
https://kgbook.org 101/150
2/15/23, 5:13 PM Knowledge Graphs
infoboxes in certain domains [Sáez and Hogan, 2018], being supported by Google [Pellissier Tanon
et al., 2016], and having been used as a data source for prominent end-user applications such as
Apple’s Siri, amongst others [Malyshev et al., 2018].
Aside from DBpedia, YAGO, Freebase and Wikidata, a number of other cross-domain knowledge
graphs have been developed down through the years. BabelNet [Navigli and Ponzetto, 2012], like
YAGO, is based on unifying WordNet and Wikipedia, but with the integration of additional knowl‐
edge graphs such as Wikidata, and a focus on creating a knowledge graph of multilingual lexical
forms (organised into multilingual synsets) by transforming lexicographic resources such as
Wiktionary and OmegaWiki into knowledge graphs. Compared to other knowledge graphs, lexi‐
calised knowledge graphs such as BabelNet bring together the encyclopedic information found in
Wikipedia with the lexicographic information usually found in monolingual and bilingual dictionar‐
ies. The Cyc project [Lenat, 1995] aims to encode common-sense knowledge in a machine-readable
way, where over 900 person-years of effort [Matuszek et al., 2006] have, since 1986, gone into the
Table of Contents
creation of 2.2 million facts and rules. Though Cyc is proprietary, an open subset called OpenCyc has
been published, where we refer to the comparison by Färber et al. [2018] of DBpedia, Freebase,
OpenCyc, and YAGO for further details. The Never Ending Language Learning (NELL)
project [Mitchell et al., 2018] has, since 2010, extracted a graph of 120 million edges from the text
of web pages using OIE methods (see Chapter 6). Each such open knowledge graph applies different
combinations of the languages and techniques discussed in this book over different sources with dif‐
fering results.
Open knowledge graphs have been published in a variety of specific domains. Schmachtenberg et al.
[2014] identify the most prominent domains in the context of Linked Data as follows: media, relating
to news, television, radio, etc. (e.g., the BBC World Service Archive [Raimond et al., 2014]); govern‐
ment, relating to the publication of data for transparency and development (e.g., by the
U.S. [Hendler et al., 2012] and U.K. [Shadbolt and O'Hara, 2013] governments); publications, relating
to academic literature in various disciplines (e.g., OpenCitations [Peroni et al., 2017],
SciGraph [Iana et al., 2019], Microsoft Academic Knowledge Graph [Färber, 2019]); geographic, re‐
lating to places and regions of interest (e.g., LinkedGeoData [Stadler et al., 2012]); life sciences, relat‐
ing to proteins, genes, drugs, diseases, etc. (e.g., Bio2RDF [Callahan et al., 2013]); and user-generated
content, relating to reviews, open source projects, etc. (e.g., Revyu [Heath and Motta, 2008]). Open
knowledge graphs have also been published in other domains, including cultural heritage [Hyvönen et
al., 2009], music [Raimond et al., 2009], law [Montiel-Ponsoda et al., 2017], theology [Sherif and
Ngonga Ngomo, 2015], and even tourism [Lu et al., 2016, Kärle et al., 2018, Maturana et al., 2018,
Zhang et al., 2019]. The envisaged applications for such knowledge graphs are as varied as the do‐
mains from which they emanate, but often relate to integration [Raimond et al., 2009, Callahan et
al., 2013], recommendation [Raimond et al., 2009, Lu et al., 2016], transparency [Hendler et al.,
2012, Shadbolt and O'Hara, 2013], archiving [Hyvönen et al., 2009, Raimond et al., 2014], decen‐
tralisation [Heath and Motta, 2008], multilingual support [Sherif and Ngonga Ngomo, 2015], regula‐
tory compliance [Montiel-Ponsoda et al., 2017], etc.
A variety of companies have announced the creation of proprietary “enterprise knowledge graphs”
with a variety of goals in mind, which include: improving search capabilities [Singhal, 2012,
Shrivastava, 2017, Krishnan, 2018, Chang, 2018, Hamad et al., 2018], providing user recommenda‐
tions [Chang, 2018, Hamad et al., 2018], implementing conversational/personal agents [Pittman et
al., 2017], enhancing targeted advertising [He et al., 2016], empowering business analytics [He et
al., 2016], connecting users [He et al., 2016, Noy et al., 2019], extending multilingual support [He
et al., 2016], facilitating research and discovery [Bendtsen and Petrovski, 2019], assessing and miti‐
Give us feedback!
https://kgbook.org 102/150
2/15/23, 5:13 PM Knowledge Graphs
gating risk [Tobin, 2017, Dalgliesh, 2016], tracking news events [Meij, 2019], and increasing trans‐
port automation [Henson et al., 2019], amongst (many) others. Though highly diverse, these enter‐
prise knowledge graphs do follow some high-level trends, as reflected in the discussion by Noy et al.
[2019]: (1) data are typically integrated into the knowledge graph from a variety of both external
and internal sources (often involving text); (2) the enterprise knowledge graph is often very large,
with millions or even billions of nodes and edges, posing challenges in terms of scalability; (3) re‐
finement of the initial knowledge graph – adding new links, consolidating duplicate entities, etc. – is
important to improve quality; (4) techniques to keep the knowledge graph up-to-date with the do‐
main are often crucial; (5) a mix of ontological and machine learning representations are often com‐
bined or used in different situations in order to draw conclusions from the enterprise knowledge
graph; (6) the ontologies used tend to be lightweight, often simple taxonomies representing a hierar‐
chy of classes or concepts. We now discuss the main industries in which enterprise knowledge graphs
have been deployed.
Web search engines have traditionally focused on matching a query string with sub-strings in web
documents. The Google Knowledge Graph [Singhal, 2012, Noy et al., 2019] rather promoted a par‐
adigm of “things not strings” – analogous to semantic search [Guha et al., 2003] – where the search
engine would now try to identify the entities that a particular search may be expressing interest in.
The knowledge graph itself describes these entities and how they interrelate. One of the main user-
facing applications of the Google Knowledge Graph is the “Knowledge Panel”, which presents a pane
on the right-hand side of (some) search results describing the principal entity that the search appears
to be seeking, including some images, attribute–value pairs, and a list of related entities that users
also search for. The Google Knowledge Graph was key to popularising the modern usage of the
phrase “knowledge graph” (see Appendix A). Other major search engines, such as Microsoft Bing38
[Shrivastava, 2017], would later announce knowledge graphs along similar lines.
note 38 Microsoft’s Knowledge Graph was previo
ing in Japanese).
10.2.2 Commerce
Enterprise knowledge graphs have also been announced by companies that are principally concerned
with selling or renting goods and services. A prominent example of such a knowledge graph is that
used by Amazon [Krishnan, 2018, Dong, 2019], which describes the products on sale in their online
marketplace. One of the main stated goals of this knowledge graph is to enable more advanced (se‐
mantic) search features for products, as well as to improve product recommendations to users of its
online marketplace. Another knowledge graph for commerce was announced by eBay [Pittman et al.,
2017], which encodes product descriptions and shopping behaviour patterns, and is used to power
conversational agents that help users to find relevant products through a natural language interface.
Airbnb [Chang, 2018] has also described a knowledge graph that encodes accommodation for rent,
places, events, experiences, neighbourhoods, users, tags, etc., on top of which a taxonomic schema is
defined. This knowledge graph is used to offer potential clients recommendations of attractions,
events, and activities available in the neighbourhood of a particular home for rent. Uber [Hamad et
al., 2018] has similarly announced a knowledge graph focused on food and restaurants for their
“Uber Eats” delivery service. The goals are again to offer semantic search features and recommenda‐
tions to users who are uncertain of precisely what kind of food they are looking for.
Enterprise knowledge graphs have also emerged in the context of social networking services.
Facebook [Noy et al., 2019] has gathered together a knowledge graph describing not only social data
about users, but also the entities they are interested in, including celebrities, places, movies, music,
etc., in order to connect people, understand their interests, and provide recommendations.
LinkedIn [He et al., 2016] announced a knowledge graph containing users, jobs, skills, companies,
places, schools, etc., on top of which a taxonomic schema is defined. The knowledge graph is used to
provide multilingual translations of important concepts, to improve targeted advertising, to provide
Give
advanced features for job search and people search, and likewise to provide recommendations us feedback!
https://kgbook.org 103/150
2/15/23, 5:13 PM Knowledge Graphs
matching jobs to people (and vice versa). Another knowledge graph has been created by
Pinterest [Gon\c{c}alves et al., 2019], describing users and their interests, the latter being organised
into a taxonomy. The main use-cases for the knowledge graph are to help users to more easily find
content of interest to them, as well as to enhance revenue through targeted advertisements.
10.2.4 Finance
The financial sector has also seen deployment of enterprise knowledge graphs. Amongst these,
Bloomberg [Meij, 2019] has proposed a knowledge graph that powers financial data analytics, in‐
cluding sentiment analysis for companies based on current news reports and tweets, a question an‐
swering service, as well as detecting emerging events that may affect stock values. Thomson Reuters
(Refinitiv) [Tobin, 2017] has likewise announced a knowledge graph encoding “the financial ecosys‐
tem” of people, organisations, equity instruments, industry classifications, joint ventures and al‐
liances, supply chains, etc., using a taxonomic schema to organise these entities. Some of the ap‐
plications they mention for the knowledge graph include supply chain monitoring, risk assessment,
and investment research. Knowledge graphs have also been used for deductive reasoning, with Banca
Table of Contents
d’Italia [Bellomarini et al., 2019] using rule-based reasoning to determine, for example, the percent‐
age of ownership of a company by various stakeholders. Other companies exploring financial knowl‐
edge graphs include Accenture [Okorafor and Ray, 2019], Capital One [Branum and Sehon, 2019],
Wells Fargo [Newman, 2019], amongst various others.
Enterprises have also been actively developing knowledge graphs to enable novel applications in a
variety of other industries, including: healthcare, where IBM are exploring use-cases for drug discov‐
ery [Noy et al., 2019] and information extraction from package inserts [Gentile et al., 2019], while
AstraZeneca [Bendtsen and Petrovski, 2019] are using a knowledge graph to advance genomics re‐
search and disease understanding; transport, where Bosch are exploring a knowledge graph of scenes
and locations for driving automation [Henson et al., 2019]; oil & gas, where Maana [Dalgliesh, 2016]
are using knowledge graphs to perform data integration for risk mitigation regarding oil wells and
drilling; and more besides.
Chapter 11
In this final chapter, we provide some concluding remarks, and further offer some insights on poten‐
tial future directions for research on knowledge graphs.
Concluding remarks. Knowledge graphs have garnered significant attention not only from diverse
organisations and industries, but also diverse research communities. This attention is due, in no
Give us feedback!
https://kgbook.org 104/150
2/15/23, 5:13 PM Knowledge Graphs
small part, to the ubiquitous nature of the problem that knowledge graphs address: integrating and
extracting value from diverse sources of data at large scale, be it in the context of a particular organ‐
isation, community, or more general collections of human knowledge. The key insight of knowledge
graphs is that graphs provide a simple, flexible, intuitive and yet powerful abstraction for represent‐
ing and integrating diverse data at large scale. This insight is far from new (see Appendix A), but
rather has finally come of age with the advent of knowledge graphs. Graphs have long been used to
represent data and knowledge in areas such as Graph Algorithms and Theory, Graph Databases,
Information Extraction, Knowledge Representation, Machine Learning, the Semantic Web, and more
besides. The advances in these areas can now be unified and applied for knowledge graphs.
Thus, the decision to model data as a graph opens up a “tool-box” of languages, techniques and sys‐
tems – stemming from diverse areas – that can be deployed in order to integrate and extract value
from data at large scale, as follows:
A variety of graph query languages are now available that (unlike other NoSQL alternatives) are
fully-featured, supporting not only the relational algebra, but also novel features such as navi‐
gational queries that can match paths of arbitrary length. A broad selection of graph databases
Table of Contents
and user interfaces supporting these query languages are now also available.
Though graphs do not depend on a detailed (relational-like) schema to represent data, various
notions of graph schemata have been proposed in order to validate, summarise and define the
semantics of graphs.
Contextual frameworks for graphs can be used to represent and reason about the scope of truth
of knowledge in the graph – relating to the time, space, provenance, confidence level, etc., for
which something is held true – including various alternatives for reification, annotated graph
frameworks, etc.
Deductive forms of reasoning can be enabled over graphs using ontologies and/or rules, which
can not only encode a machine-readable consensus about the meaning of the graph, but also
provide automated access to implicit knowledge entailed by a graph through materialisation or
query rewriting.
Graph algorithms, such as centrality measures, community detection, clustering, etc., can be ap‐
plied on the data to gain insights about influential entities or edges, close-knit sub-graphs of
entities, and more besides, with graph parallel frameworks capable of applying such algorithms
at large scale.
Recent and continual advances in knowledge graph embeddings and graph neural networks have
now opened up new possibilities for applying machine learning natively over graphs in the
context of diverse tasks, including classification, question answering, recommendations, and
more besides.
Rule and axiom mining techniques allow for extracting formal, declarative hypotheses from a
knowledge graph that encode high-level patterns and can be applied to derive new knowledge
in a deductive, explainable manner.
Graph-based information extraction can be applied to extract and/or enrich a knowledge graph
from legacy sources of text and semi-structured data, while graph-based mapping languages facil‐
itate integrating diverse sources of legacy structured data into the knowledge graph.
Tools, techniques and methodologies for ontology engineering and ontology learning can further
guide the – potentially collaborative – creation of an ontology for the knowledge graph, encod‐
ing a consensus about its semantics, and enabling access to implicit knowledge through deduc‐
tive reasoning.
Quality dimensions and metrics for knowledge graphs allow for systematically assessing the readi‐
ness of the knowledge graph for its envisaged applications, in both a qualitative and quantita‐
tive manner, where a variety of tools and frameworks are available to help perform such
assessments.
Knowledge graphs that have been integrated from diverse sources are likely to be incomplete,
or to encode incorrect data, where techniques and tools for knowledge graph refinement facilitate
the automated completion and correction of knowledge graphs, thus improving its overall qual‐
ity and usefulness.
For the purposes of publishing open knowledge graphs, principles & best practices and access pro‐
Give
tocols, as well as techniques for linking, licensing, access & usage control, encryption and anonymi‐ us feedback!
https://kgbook.org 105/150
2/15/23, 5:13 PM Knowledge Graphs
sation, can be leveraged to maximise their potential impact on society in an ethical way.
As we have discussed in Chapter 10, the various components of this “knowledge graph tool-box” can
already be found deployed in practice, having been applied – to varying degrees – in the context of
numerous open and enterprise knowledge graphs. As adoption of knowledge graphs continues, work
will also continue on improving and combining these tools, as well as on developing novel tools that
help to better integrate and extract value from diverse sources of data at large scale.
Future directions. Research on knowledge graphs involves a confluence of techniques from different
research areas with the common objective of maximising the knowledge – and thus value – that can
be distilled from diverse sources at large scale using a graph-based data abstraction [Hogan, 2020a].
In the intersection of data graphs and deductive knowledge, we emphasise emerging topics such as
formal semantics for property graphs, with languages that can take into account the meaning of labels
and property–value pairs on nodes and edges [Krötzsch et al., 2018]; and reasoning and querying over
contextual data, in order to derive conclusions and results valid in a particular setting [Serafini and
Homola, 2012, Zimmermann et al., 2012, Schuetz et al., 2021]. In the intersection of data graphs
Table of Contents
and inductive knowledge, we highlight topics such as similarity-based query relaxation, allowing to
find approximate answers to exact queries based on numerical representations (e.g.,
embeddings) [Wang et al., 2018]; shape induction, in order to learn and formalise inherent patterns
in the knowledge graph as constraints [Mihindukulasooriya et al., 2018]; and contextual knowledge
graph embeddings that provide numeric representations of nodes and edges that vary with time, place,
etc. [Kazemi et al., 2019]. In the intersection of deductive and inductive knowledge, we mention the
topics of entailment-aware knowledge graph embeddings [Guo et al., 2016, Demeester et al., 2016], that
incorporate rules and/or ontologies when computing plausibility; expressive graph neural networks
proven capable of complex classification analogous to expressive ontology languages [Barceló et al.,
2020]; as well as further advances on rule and axiom mining, allowing to extract symbolic, deductive
representations from the knowledge graphs [Galárraga et al., 2015, Bühmann et al., 2016]. Further
challenges arise when considering the creation, enrichment, refinement and publication of knowl‐
edge graphs, which call for further works on topics such as automated quality assessment (and repair),
distantly-supervised extraction frameworks, efficient access protocols, and anonymisation, to name but a
few.
Aside from specific topics, more general challenges for knowledge graphs include scalability, particu‐
larly for deductive and inductive reasoning; quality, not only in terms of data, but also the models in‐
duced from knowledge graphs; diversity, such as managing contextual or multi-modal data; dynamici‐
ty, considering temporal or streaming data; and finally usability, which is key to increasing adoption.
Though techniques are continuously being proposed to address these challenges, they are unlikely to
ever be completely “solved”; rather they serve as dimensions along which knowledge graphs, and
their techniques, tools, etc., will continue to mature.
Given the availability of open knowledge graphs whose quality continues to improve, as well as the
growing adoption of enterprise knowledge graphs in various industries, future research on knowl‐
edge graphs has the potential to foster key advancements in broad aspects of society. Here we have
highlighted just some examples of future research directions of importance to this pursuit.
Bibliography
Serge Abiteboul. 1997. Querying Semi-Structured Data. In Database Theory - ICDT '97, 6th International
Conference, Delphi, Greece, January 8-10, 1997, Proceedings, Foto N. Afrati and Phokion G. Kolaitis
Give us feedback!
(Eds.). Lecture Notes in Computer Science, vol. 1186. Springer, 1–18.
https://kgbook.org 106/150
2/15/23, 5:13 PM Knowledge Graphs
Sushant Agarwal, Simon Steyskal, Franjo Antunovic, and Sabrina Kirrane. 2018. Legislative Compliance
Assessment: Framework, Model and GDPR Instantiation. In Privacy Technologies and Policy - 6th
Annual Privacy Forum, APF 2018, Barcelona, Spain, June 13-14, 2018, Revised Selected Papers, Manel
Medina, Andreas Mitrakas, Kai Rannenberg, Erich Schweighofer, and Nikolaos Tsouroulas (Eds.).
Lecture Notes in Computer Science, vol. 11079. Springer, 131–149.
Rakesh Agrawal, Tomasz Imieliński, and Arun Swami. 1993. Mining association rules between sets of
items in large databases. In Proceedings of the 1993 ACM SIGMOD International Conference on
Management of Data, Washington, DC, USA, May 26-28, 1993, Peter Buneman and Sushil Jajodia
(Eds.). ACM Press, 207–216.
Jinhyun Ahn, Dong-Hyuk Im, Jae-Hong Eom, Nansu Zong, and Hong-Gee Kim. 2015. G-Diff: A Grouping
Algorithm for RDF Change Detection on MapReduce. In Semantic Technology - 4th Joint International
Conference, JIST 2014, Chiang Mai, Thailand, November 9-11, 2014. Revised Selected Papers, Thepchai
Supnithi, Takahira Yamaguchi, Jeff Z. Pan, Vilas Wuwongse, and Marut Buranarach (Eds.). Lecture
Notes in Computer Science, vol. 8943. Springer, 230–235.
Adnan Akhter, Axel-Cyrille Ngonga Ngomo, and Muhammad Saleem. 2018. An Empirical Evaluation of
RDF Graph Partitioning Techniques. In Knowledge Engineering and Knowledge Management - 21st
Table of Contents
International Conference, EKAW 2018, Nancy, France, November 12-16, 2018, Proceedings, Catherine
Faron-Zucker, Chiara Ghidini, Amedeo Napoli, and Yannick Toussaint (Eds.). Lecture Notes in
Computer Science, vol. 11313. Springer, 3–18.
Keith Alexander, Richard Cyganiak, Michael Hausenblas, and Jun Zhao. 2009. Describing Linked
Datasets. In Proceedings of the WWW2009 Workshop on Linked Data on the Web, LDOW 2009, Madrid,
Spain, April 20, 2009, Christian Bizer, Tom Heath, Tim Berners-Lee, and Michael Hausenblas (Eds.).
CEUR Workshop Proceedings, vol. 538. Sun SITE Central Europe (CEUR). 10 pages.
Renzo Angles and Claudio Gutierrez. 2008. Survey of graph database models. ACM Computing Surveys
40(1), 1:1–1:39.
Renzo Angles, Marcelo Arenas, Pablo Barceló, Aidan Hogan, Juan L. Reutter, and Domagoj Vrgoč. 2017.
Foundations of Modern Query Languages for Graph Databases. ACM Computing Surveys 50(5), 68:1–
68:40.
Renzo Angles, Marcelo Arenas, Pablo Barceló, Peter A. Boncz, George H. L. Fletcher, Claudio Gutierrez,
Tobias Lindaaker, Marcus Paradies, Stefan Plantikow, Juan F. Sequeda, Oskar van Rest, and Hannes
Voigt. 2018. G-CORE: A Core for Future Graph Query Languages. In Proceedings of the 2018
International Conference on Management of Data, SIGMOD Conference 2018, Houston, TX, USA, June
10-15, 2018, Gautam Das, Christopher M. Jermaine, and Philip A. Bernstein (Eds.). ACM Press,
1421–1432.
Renzo Angles, Harsh Thakkar, and Dominik Tomaszuk. 2019. RDF and Property Graphs Interoperability:
Status and Issues. In Proceedings of the 13th Alberto Mendelzon International Workshop on Foundations
of Data Management, Asunción, Paraguay, June 3-7, 2019, Aidan Hogan and Tova Milo (Eds.). CEUR
Workshop Proceedings, vol. 2369. Sun SITE Central Europe (CEUR). 11 pages.
Renzo Angles. 2018. The Property Graph Database Model. In Proceedings of the 12th Alberto Mendelzon
International Workshop on Foundations of Data Management, Cali, Colombia, May 21–25, 2018, Dan
Olteanu and Barbara Poblete (Eds.). CEUR Workshop Proceedings, vol. 2100. Sun SITE Central
Europe (CEUR). 10 pages.
Mikel Egaña Aranguren, Erick Antezana, Martin Kuiper, and Robert Stevens. 2008. Ontology Design
Patterns for bio-ontologies: a case study on the Cell Cycle Ontology. BMC Bioinformatics 9(5), pS1.
Marcelo Arenas, Alexandre Bertails, Eric Prud’hommeaux, and Juan Sequeda. 2012. A Direct Mapping of
Relational Data to RDF, W3C Recommendation 27 September 2012. W3C Recommendation. World
Wide Web Consortium. September 27, 2012.
Marcelo Arenas, Bernardo Cuenca Grau, Evgeny Kharlamov, Sarunas Marciuska, and Dmitriy
Zheleznyakov. 2016. Faceted search over RDF-based knowledge graphs. Journal of Web Semantics 37-
38, 55–74.
Alessandro Artale, Diego Calvanese, Roman Kontchakov, and Michael Zakharyaschev. 2009. The DL-Lite
Family and Relations. Journal of Artificial Intelligence Research 36, 1–69.
Sören Auer, Christian Bizer, Georgi Kobilarov, Jens Lehmann, Richard Cyganiak, and Zachary Ives. 2007.
DBpedia: A Nucleus for a Web of Open Data. In The Semantic Web, 6th International Semantic Web
Conference, 2nd Asian Semantic Web Conference, ISWC 2007 + ASWC 2007, Busan, Korea, November
11-15, 2007, Karl Aberer, Key-Sun Choi, Natasha Fridman Noy, Dean Allemang, Kyung-Il Lee,
Give us feedback!
https://kgbook.org 107/150
2/15/23, 5:13 PM Knowledge Graphs
Lyndon J. B. Nixon, Jennifer Golbeck, Peter Mika, Diana Maynard, Riichiro Mizoguchi, Guus
Schreiber, and Philippe Cudré-Mauroux (Eds.). Lecture Notes in Computer Science, vol. 4825.
Springer, 722–735.
Franz Baader, Ian Horrocks, Carsten Lutz, and Ulrike Sattler. 2017. An Introduction to Description Logic.
Cambridge University Press.
Nguyen Bach and Sameer Badaskar. 2007. A Review of Relation Extraction. Carnegie Mellon University.
Ricardo Baeza-Yates. 2018. Bias on the Web. Communications of the ACM 61(6), 54–61.
Collin F. Baker, Charles J. Fillmore, and John B. Lowe. 1998. The Berkeley FrameNet Project. In 36th
Annual Meeting of the Association for Computational Linguistics and 17th International Conference on
Computational Linguistics, COLING-ACL'98, August 10-14, 1998, Université de Montréal, Montréal,
Quebec, Canada. Proceedings of the Conference, Christian Boitet and Pete Whitelock (Eds.). Morgan
Kaufmann / The Association for Computational Linguistics, 86–90.
René Ronald Bakker. 1987. Knowledge Graphs: Representation and Structuring of Scientific Knowledge. Ph.D.
dissertation. University of Twente, Enschede, Netherlands.
Ivana Balazevic, Carl Allen, and Timothy M. Hospedales. 2019. Multi-relational Poincaré Graph
Embeddings. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural
Table of Contents
Information Processing Systems 2019, NeurIPS 2019, 8–14 December 2019, Vancouver, BC, Canada,
Hanna M. Wallach, Hugo Larochelle, Alina Beygelzimer, Florence d’Alché-Buc, Emily B. Fox, and
Roman Garnett (Eds.), 4465–4475.
Ivana Balazevic, Carl Allen, and Timothy M. Hospedales. 2019. TuckER: Tensor Factorization for
Knowledge Graph Completion. In Proceedings of the 2019 Conference on Empirical Methods in Natural
Language Processing and the 9th International Joint Conference on Natural Language Processing, EMNLP-
IJCNLP 2019, Hong Kong, China, November 3-7, 2019, Kentaro Inui, Jing Jiang, Vincent Ng, and
Xiaojun Wan (Eds.). The Association for Computational Linguistics, 5184–5193.
Ivana Balazevic, Carl Allen, and Timothy M. Hospedales. 2019. Hypernetwork Knowledge Graph
Embeddings. In Artificial Neural Networks and Machine Learning - ICANN 2019 - 28th International
Conference on Artificial Neural Networks, Munich, Germany, September 17-19, 2019, Proceedings -
Workshop and Special Sessions, Igor V. Tetko, Vera Kurková, Pavel Karpov, and Fabian J. Theis (Eds.).
Lecture Notes in Computer Science, vol. 11731. Springer, 553–565.
Michele Banko, Michael J. Cafarella, Stephen Soderland, Matthew Broadhead, and Oren Etzioni. 2007.
Open Information Extraction from the Web. In IJCAI 2007, Proceedings of the 20th International Joint
Conference on Artificial Intelligence, Hyderabad, India, January 6-12, 2007, Manuela M. Veloso (Ed.).
AAAI Press, 2670–2676.
Pablo Barceló, Egor V. Kostylev, Mikael Monet, Jorge Peréz, Juan Reutter, and Juan Pablo Silva. 2020.
The Logical Expressiveness of Graph Neural Networks. In 8th International Conference on Learning
Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26–30, 2020. OpenReview.net. 20 pages.
Hannah Bast and Björn Buchhold. 2013. An index for efficient semantic full-text search. In 22nd ACM
International Conference on Information and Knowledge Management, CIKM'13, San Francisco, CA, USA,
October 27 - November 1, 2013, Qi He, Arun Iyengar, Wolfgang Nejdl, Jian Pei, and Rajeev Rastogi
(Eds.). ACM Press, 369–378.
Carlo Batini and Monica Scannapieco. 2016. Data and Information Quality - Dimensions, Principles and
Techniques. Data-Centric Systems and Applications. Springer.
Carlo Batini, Anisa Rula, Monica Scannapieco, and Gianluigi Viscusi. 2015. From Data Quality to Big
Data Quality. Journal of Database Management 26(1), 60–82.
Luigi Bellomarini, Emanuel Sallinger, and Georg Gottlob. 2018. The Vadalog System: Datalog-based
Reasoning for Knowledge Graphs. Proceedings of the VLDB Endowment 11(9), 975–987.
Luigi Bellomarini, Daniele Fakhoury, Georg Gottlob, and Emanuel Sallinger. 2019. Knowledge Graphs
and Enterprise AI: The Promise of an Enabling Technology. In 35th IEEE International Conference on
Data Engineering, ICDE 2019, Macao, China, April 8-11, 2019. IEEE Computer Society, 26–37.
Claus Bendtsen and Slavé Petrovski. 2019. How data and AI are helping unlock the secrets of disease.
AstraZeneca Blog. November 1, 2019.
Michael K. Bergman. 2019. A Common Sense View of Knowledge Graphs. Adaptive Information,
Adaptive Innovation, Adaptive Infrastructure Blog. July 1, 2019.
Tim Berners-Lee and Dan Connolly. 2011. Notation3 (N3): A readable RDF syntax. W3C Team Submission.
World Wide Web Consortium. March 28, 2011.
Give us feedback!
https://kgbook.org 108/150
2/15/23, 5:13 PM Knowledge Graphs
Tim Berners-Lee, James Hendler, and Ora Lassila. 2001. The Semantic Web. Scientific American 284(5),
34–43.
Tim Berners-Lee. 2006. Linked Data. W3C Design Issues. July, 2006.
Anant P. Bhardwaj, Souvik Bhattacherjee, Amit Chavan, Amol Deshpande, Aaron J. Elmore, Samuel
Madden, and Aditya G. Parameswaran. 2015. DataHub: Collaborative Data Science & Dataset
Version Management at Scale. In CIDR 2015, Seventh Biennial Conference on Innovative Data Systems
Research, Asilomar, CA, USA, January 4-7, 2015, Online Proceedings. www.cidrdb.org. 7 pages.
Daniel M. Bikel, Richard M. Schwartz, and Ralph M. Weischedel. 1999. An Algorithm that Learns What’s
in a Name. Machine Learning 34(1–3), 211–231.
Stefan Bischof, Stefan Decker, Thomas Krennwallner, Nuno Lopes, and Axel Polleres. 2012. Mapping be‐
tween RDF and XML with XSPARQL. Journal of Web Semantics 1(3), 147–185.
Christian Bizer, Jens Lehmann, Georgi Kobilarov, Sören Auer, Christian Becker, Richard Cyganiak, and
Sebastian Hellmann. 2009. DBpedia - A crystallization point for the Web of Data. Journal of Web
Semantics 7(3), 154–165.
Eva Blomqvist and Kurt Sandkuhl. 2005. Patterns in Ontology Engineering: Classification of Ontology
Patterns. In ICEIS 2005, Proceedings of the Seventh International Conference on Enterprise Information
Table of Contents
Systems, Miami, USA, May 25-28, 2005, Chin-Sheng Chen, Joaquim Filipe, Isabel Seruca, and José
Cordeiro (Eds.), vol. 3, 413–416.
Eva Blomqvist, Azam Seil Sepour, and Valentina Presutti. 2012. Ontology Testing - Methodology and
Tool. In Knowledge Engineering and Knowledge Management - 18th International Conference, EKAW
2012, Galway City, Ireland, October 8-12, 2012. Proceedings, Annette ten Teije, Johanna Völker,
Siegfried Handschuh, Heiner Stuckenschmidt, Mathieu d’Aquin, Andriy Nikolov, Nathalie Aussenac-
Gilles, and Nathalie Hernandez (Eds.). Lecture Notes in Computer Science, vol. 7603. Springer, 216–
226.
Eva Blomqvist, Karl Hammar, and Valentina Presutti. 2016. Engineering Ontologies with Patterns – The
eXtreme Design Methodology. In Ontology Engineering with Ontology Design Patterns - Foundations and
Applications, Pascal Hitzler, Aldo Gangemi, Krzysztof Janowicz, Adila Krisnadhi, and Valentina
Presutti (Eds.). Studies on the Semantic Web, vol. 25. IOS Press, 23–50.
Kurt Bollacker, Patrick Tufts, Tomi Pierce, and Robert Cook. 2007. A platform for scalable, collaborative,
structured information integration. In Intl. Workshop on Information Integration on the Web (IIWeb'07),
Ullas Nambiar and Zaiqing Nie (Eds.). 6 pages.
Kurt Bollacker, Robert Cook, and Patrick Tufts. 2007. Freebase: A Shared Database of Structured General
Human Knowledge. In Proceedings of the Twenty-Second AAAI Conference on Artificial Intelligence, July
22-26, 2007, Vancouver, British Columbia, Canada. AAAI Press, 1962–1963.
Piero A. Bonatti and Sabrina Kirrane. 2019. Big Data and Analytics in the Age of the GDPR. In 2019 IEEE
International Congress on Big Data, BigData Congress 2019, Milan, Italy, July 8-13, 2019, Elisa Bertino,
Carl K. Chang, Peter Chen, Ernesto Damiani, Michael Goul, and Katsunori Oyama (Eds.). IEEE
Computer Society, 7–16.
Piero A. Bonatti, Aidan Hogan, Axel Polleres, and Luigi Sauro. 2011. Robust and scalable Linked Data
reasoning incorporating provenance and trust annotations. Journal of Web Semantics 9(2), 165–201.
Piero Andrea Bonatti, Stefan Decker, Axel Polleres, and Valentina Presutti. 2018. Knowledge Graphs:
New Directions for Knowledge Representation on the Semantic Web (Dagstuhl Seminar 18371).
Dagstuhl Reports 8(9), 29–111.
Piero Bonatti, Sabrina Kirrane, Iliana Mineva Petrova, Luigi Sauro, and Eva Schlehahn. 2019. The
SPECIAL Usage Policy Language, version 1.0. Unofficial Draft. Vienna University of Economics and
Business. December 31, 2019.
Iovka Boneva, Jose Emilio Labra Gayo, and Eric G. Prud’hommeaux. 2017. Semantics and Validation of
Shapes Schemas for RDF. In The Semantic Web - ISWC 2017 - 16th International Semantic Web
Conference, Vienna, Austria, October 21-25, 2017, Proceedings, Part I, Claudia d’Amato, Miriam
Fernández, Valentina A. M. Tamma, Freddy Lécué, Philippe Cudré-Mauroux, Juan F. Sequeda,
Christoph Lange, and Jeff Heflin (Eds.). Lecture Notes in Computer Science, vol. 10587. Springer,
104–120.
Iovka Boneva, Jérémie Dusart, Daniel Fernández-álvarez, and José Emilio Labra Gayo. 2019. Shape
Designer for ShEx and SHACL constraints. In Proceedings of the ISWC 2019 Satellite Tracks (Posters &
Demonstrations, Industry, and Outrageous Ideas) co-located with 18th International Semantic Web
Conference (ISWC 2019), Auckland, New Zealand, October 26-30, 2019, Mari Carmen Suárez-Figueroa,
Give us feedback!
https://kgbook.org 109/150
2/15/23, 5:13 PM Knowledge Graphs
Gong Cheng, Anna Lisa Gentile, Christophe Guéret, C. Maria Keet, and Abraham Bernstein (Eds.).
CEUR Workshop Proceedings, vol. 2456. Sun SITE Central Europe (CEUR), 269–272.
Angela Bonifati, Wim Martens, and Thomas Timm. 2017. An Analytical Study of Large SPARQL Query
Logs. Proceedings of the VLDB Endowment 11(2), 149–161.
Antoine Bordes, Nicolas Usunier, Alberto García-Durán, Jason Weston, and Oksana Yakhnenko. 2013.
Translating Embeddings for Modeling Multi-relational Data. In Advances in Neural Information
Processing Systems 26: 27th Annual Conference on Neural Information Processing Systems 2013.
Proceedings of a meeting held December 5-8, 2013, Lake Tahoe, Nevada, United States, Christopher J. C.
Burges, Léon Bottou, Zoubin Ghahramani, and Kilian Q. Weinberger (Eds.), 2787–2795.
Gerlof Bouma. 2009. Normalized (Pointwise) Mutual Information in Collocation Extraction. In Von der
Form zur Bedeutung: Texte automatisch verarbeiten - From Form to Meaning: Processing Texts
Automatically, Proceedings of the Biennial GSCL Conference 2009, Christian Chiarcos, Richard Eckart de
Castilho, and Manfred Stede (Eds.). Gunter Narr Verlag, 31–40.
Ronald J. Brachman and Hector J. Levesque. 1986. The Knowledge Level of a KBMS. In On Knowledge
Base Management Systems: Integrating Artificial Intelligence and Database Technologies, Book resulting
from the Islamorada Workshop 1985 (Islamorada, FL, USA), Michael L. Brodie and John Mylopoulos
Table of Contents
https://kgbook.org 110/150
2/15/23, 5:13 PM Knowledge Graphs
Michael J. Cafarella, Alon Y. Halevy, Daisy Zhe Wang, Eugene Wu, and Yang Zhang. 2008. WebTables:
exploring the power of tables on the web. Proceedings of the VLDB Endowment 1(1), 538–549.
HongYun Cai, Vincent W. Zheng, and Kevin Chen-Chuan Chang. 2018. A Comprehensive Survey of Graph
Embedding: Problems, Techniques, and Applications. IEEE Transactions on Knowledge and Data
Engineering 30(9), 1616–1637.
Alison Callahan, Jose Cruz-Toledo, Peter Ansell, and Michel Dumontier. 2013. Bio2RDF Release 2:
Improved Coverage, Interoperability and Provenance of Life Science Linked Data. In The Semantic
Web: Semantics and Big Data, 10th International Conference, ESWC 2013, Montpellier, France, May 26-
30, 2013. Proceedings, Philipp Cimiano, Óscar Corcho, Valentina Presutti, Laura Hollink, and
Sebastian Rudolph (Eds.). Lecture Notes in Computer Science, vol. 7882. Springer, 200–212.
Nicholas J. Car, Paul J. Box, and Ashley Sommer. 2019. The Location Index: A Semantic Web Spatial
Data Infrastructure. In The Semantic Web - 16th International Conference, ESWC 2019, Portorož,
Slovenia, June 2-6, 2019, Proceedings, Pascal Hitzler, Miriam Fernández, Krzysztof Janowicz,
Amrapali Zaveri, Alasdair J. G. Gray, Vanessa López, Armin Haller, and Karl Hammar (Eds.). Lecture
Notes in Computer Science, vol. 11503. Springer, 543–557.
Šejla Čebirić, François Goasdoué, Haridimos Kondylakis, Dimitris Kotzinos, Ioana Manolescu, Georgia
Table of Contents
Troullinou, and Mussab Zneika. 2019. Summarizing semantic graphs: a survey. The Very Large Data
Base Journal 28(3), 295–327.
Stefano Ceri, Georg Gottlob, and Letizia Tanca. 1989. What you Always Wanted to Know About Datalog
(And Never Dared to Ask). IEEE Transactions on Knowledge and Data Engineering 1(1), 146–166.
Spencer Chang. 2018. Scaling Knowledge Access and Retrieval at Airbnb. AirBnB Medium Blog.
September 4, 2018.
Laura Chiticariu, Yunyao Li, and Frederick R. Reiss. 2013. Rule-Based Information Extraction is Dead!
Long Live Rule-Based Information Extraction Systems!. In Proceedings of the 2013 Conference on
Empirical Methods in Natural Language Processing, EMNLP 2013, 18-21 October 2013, Grand Hyatt
Seattle, Seattle, Washington, USA, A meeting of SIGDAT, a Special Interest Group of the ACL, David
Yarowsky, Timothy Baldwin, Anna Korhonen, Karen Livescu, and Steven Bethard (Eds.). The
Association for Computational Linguistics, 827–832.
Laura Chiticariu, Marina Danilevsky, Yunyao Li, Frederick Reiss, and Huaiyu Zhu. 2018. SystemT:
Declarative Text Understanding for Enterprise. In Proceedings of the 2018 Conference of the North
American Chapter of the Association for Computational Linguistics: Human Language Technologies,
NAACL-HLT 2018, New Orleans, Louisiana, USA, June 1-6, 2018, Volume 3 (Industry Papers), Srinivas
Bangalore, Jennifer Chu-Carroll, and Yunyao Li (Eds.). The Association for Computational
Linguistics, 76–83.
Giovanni Luca Ciampaglia, Prashant Shiralkar, Luis M Rocha, Johan Bollen, Filippo Menczer, and
Alessandro Flammini. 2015. Computational fact checking from knowledge networks. PLOS One
10(6), pe0128193.
Aaron Clemmer and Stephen Davies. 2011. Smeagol: A "Specific-to-General" Semantic Web Query
Interface Paradigm for Novices. In Database and Expert Systems Applications - 22nd International
Conference, DEXA 2011, Toulouse, France, August 29 - September 2, 2011. Proceedings, Part I,
Abdelkader Hameurlain, Stephen W. Liddle, Klaus-Dieter Schewe, and Xiaofang Zhou (Eds.).
Springer, 288–302.
Michael Cochez, Petar Ristoski, Simone Paolo Ponzetto, and Heiko Paulheim. 2017. Biased Graph Walks
for RDF Graph Embeddings. In Proceedings of the 7th International Conference on Web Intelligence,
Mining and Semantics, WIMS 2017, Amantea, Italy, June 19-22, 2017, Rajendra Akerkar, Alfredo
Cuzzocrea, Jannong Cao, and Mohand-Said Hacid (Eds.). ACM Press, 21:1–21:12.
Michael Cochez, Petar Ristoski, Simone Paolo Ponzetto, and Heiko Paulheim. 2017. Global RDF Vector
Space Embeddings. In The Semantic Web - ISWC 2017 - 16th International Semantic Web Conference,
Vienna, Austria, October 21-25, 2017, Proceedings, Part I, Claudia d’Amato, Miriam Fernández,
Valentina A. M. Tamma, Freddy Lécué, Philippe Cudré-Mauroux, Juan F. Sequeda, Christoph Lange,
and Jeff Heflin (Eds.). Lecture Notes in Computer Science, vol. 10587. Springer, 190–207.
Diego Collarana, Mikhail Galkin, Christoph Lange, Irlán Grangel-González, Maria-Esther Vidal, and Sören
Auer. 2016. FuhSen: A Federated Hybrid Search Engine for Building a Knowledge Graph On-Demand
(Short Paper). In On the Move to Meaningful Internet Systems: OTM 2016 Conferences - Confederated
International Conferences: CoopIS, C&TC, and ODBASE 2016, Rhodes, Greece, October 24-28, 2016,
Proceedings, Christophe Debruyne, Hervé Panetto, Robert Meersman, Tharam S. Dillon, eva Kühn,
Give us feedback!
https://kgbook.org 111/150
2/15/23, 5:13 PM Knowledge Graphs
Declan O’Sullivan, and Claudio Agostino Ardagna (Eds.). Lecture Notes in Computer Science,
vol. 10033. Springer, 752–761.
Michael Collins and Yoram Singer. 1999. Unsupervised Models for Named Entity Classification. In Joint
SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora,
EMNLP 1999, College Park, MD, USA, June 21-22, 1999. The Association for Computational
Linguistics. 11 pages.
Dan Connolly. 2007. Gleaning Resource Descriptions from Dialects of Languages (GRDDL), W3C
Recommendation 11 September 2007. W3C Recommendation. World Wide Web Consortium.
September 11, 2007.
Mariano P. Consens and Alberto O. Mendelzon. 1990. GraphLog: a Visual Formalism for Real Life
Recursion. In Proceedings of the Ninth ACM SIGACT-SIGMOD-SIGART Symposium on Principles of
Database Systems, April 2-4, 1990, Nashville, Tennessee, USA, Daniel J. Rosenkrantz and Yehoshua
Sagiv (Eds.). ACM Press, 404–416.
Olivier Corby and Catherine Faron-Zucker. 2010. The KGRAM Abstract Machine for Knowledge Graph
Querying. In 2010 IEEE/WIC/ACM International Conference on Web Intelligence, WI 2010, Toronto,
Canada, August 31 - September 3, 2010, Main Conference Proceedings, Jimmy Xiangji Huang, Irwin
Table of Contents
King, Vijay V. Raghavan, and Stefan Rueger (Eds.). IEEE Computer Society, 338–341.
Francesco Corcoglioniti, Marco Rospocher, and Alessio Palmero Aprosio. 2016. Frame-Based Ontology
Population with PIKES. IEEE Transactions on Knowledge and Data Engineering 28(12), 3261–3275.
Julien Corman, Juan L. Reutter, and Ognjen Savković. 2018. Semantics and Validation of Recursive
SHACL. In The Semantic Web - ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA,
USA, October 8-12, 2018, Proceedings, Part I, Denny Vrandečić, Kalina Bontcheva, Mari Carmen
Suárez-Figueroa, Valentina Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena
Simperl (Eds.). Lecture Notes in Computer Science, vol. 11136. Springer, 318–336.
Julien Corman, Fernando Florenzano, Juan L. Reutter, and Ognjen Savkovic. 2019. Validating SHACL
Constraints over a SPARQL Endpoint. In The Semantic Web - ISWC 2019 - 18th International Semantic
Web Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part I, Chiara Ghidini, Olaf
Hartig, Maria Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime
Lefrançois, and Fabien Gandon (Eds.). Lecture Notes in Computer Science, vol. 11778. Springer,
145–163.
Luca Costabello, Serena Villata, Nicolas Delaforge, and Fabien Gandon. 2012. Linked Data Access Goes
Mobile: Context-Aware Authorization for Graph Stores. In WWW2012 Workshop on Linked Data on
the Web, Lyon, France, 16 April, 2012, Christian Bizer, Tom Heath, Tim Berners-Lee, and Michael
Hausenblas (Eds.). CEUR Workshop Proceedings, vol. 937. Sun SITE Central Europe (CEUR). 8 pages.
Kino Coursey and Rada Mihalcea. 2009. Topic Identification Using Wikipedia Graph Centrality. In Human
Language Technologies: Conference of the North American Chapter of the Association of Computational
Linguistics, Proceedings, May 31 - June 5, 2009, Boulder, Colorado, USA, Short Papers, Mari Ostendorf,
Michael Collins, Shri Narayanan, Douglas W. Oard, and Lucy Vanderwende (Eds.). The Association
for Computational Linguistics, 117–120.
Simon Cox, Chris Little, Jerry R. Hobbs, and Feng Pan. 2017. Time Ontology in OWL. W3C
Recommendation / OGC 16-071r2. World Wide Web Consortium and Open Geospatial Consortium.
October 19, 2017.
Eric Crestan and Patrick Pantel. 2011. Web-scale table census and classification. In Proceedings of the
Forth International Conference on Web Search and Web Data Mining, WSDM 2011, Hong Kong, China,
February 9-12, 2011, Irwin King, Wolfgang Nejdl, and Hang Li (Eds.). ACM Press, 545–554.
Richard Cyganiak, David Wood, and Markus Lanthaler. 2014. RDF 1.1 Concepts and Abstract Syntax. W3C
Recommendation. World Wide Web Consortium. February 25, 2014.
Enrico Daga, Valentina Presutti, and Alberto Salvati. 2008. http://ontologydesignpatterns.org and
Evaluation WikiFlow. In Proceedings of the 5th Workshop on Semantic Web Applications and Perspectives
(SWAP2008), Rome, Italy, December 15-17, 2008, Aldo Gangemi, Johannes Keizer, Valentina Presutti,
and Heiko Stoermer (Eds.). CEUR Workshop Proceedings, vol. 426. Sun SITE Central Europe (CEUR),
1–11.
Jeff Dalgliesh. 2016. How the Enterprise Knowledge Graph Connects Oil and Gas Data Silos. Maana Blog.
May 20, 2016.
Claudia d’Amato, Steffen Staab, Andrea G. B. Tettamanzi, Duc Minh Tran, and Fabien L. Gandon. 2016.
Ontology enrichment by discovering multi-relational association rules from ontological knowledge
Give us feedback!
https://kgbook.org 112/150
2/15/23, 5:13 PM Knowledge Graphs
bases. In Proceedings of the 31st Annual ACM Symposium on Applied Computing, Pisa, Italy, April 4-8,
2016, Sascha Ossowski (Ed.). ACM Press, 333–338.
Claudia d’Amato, Andrea G. B. Tettamanzi, and Minh Duc Tran. 2016. Evolutionary Discovery of Multi-
relational Association Rules from Ontological Knowledge Bases. In Knowledge Engineering and
Knowledge Management - 20th International Conference, EKAW 2016, Bologna, Italy, November 19-23,
2016, Proceedings, Eva Blomqvist, Paolo Ciancarini, Francesco Poggi, and Fabio Vitali (Eds.). Lecture
Notes in Computer Science, vol. 10024. Springer, 113–128.
Fariz Darari, Werner Nutt, Giuseppe Pirrò, and Simon Razniewski. 2018. Completeness Management for
RDF Data Sources. ACM Transactions on the Web 12(3), 18:1–18:53.
Souripriya Das, Seema Sundara, and Richard Cyganiak. 2012. R2RML: RDB to RDF Mapping Language,
W3C Recommendation 27 September 2012. W3C Recommendation. World Wide Web Consortium.
September 27, 2012.
Ankur Dave, Alekh Jindal, Li Erran Li, Reynold Xin, Joseph Gonzalez, and Matei Zaharia. 2016.
GraphFrames: an integrated API for mixing graph and relational queries. In Proceedings of the Fourth
International Workshop on Graph Data Management Experiences and Systems, Redwood Shores, CA, USA,
June 24 - 24, 2016, Peter A. Boncz and Josep-Lluís Larriba-Pey (Eds.). ACM Press, 2:1–2:8.
Table of Contents
Gerard de Melo. 2015. Lexvo.org: Language-Related Information for the Linguistic Linked Data Cloud.
Semantic Web Journal 6(4), 393–400.
Luc De Raedt, Bart Vandersmissen, Marc Denecker, and Maurice Bruynooghe. 1991. A hybrid approach
to learning and its knowledge representation. In Proceedings of the third COGNITIVA symposium on At
the crossroads of artificial intelligence, cognitive science, and neuroscience, Teuvo Kohonen and Françoise
Fogelman-Soulie (Eds.). Elsevier, 109–116.
Luc De Raedt (Ed.). 2008. Logical and Relational Learning. Springer.
Marina De Vos, Sabrina Kirrane, Julian Padget, and Ken Satoh. 2019. ODRL policy modelling and compli‐
ance checking. In Rules and Reasoning - Third International Joint Conference, RuleML+RR 2019,
Bolzano, Italy, September 16-19, 2019, Proceedings, Paul Fodor, Marco Montali, Diego Calvanese, and
Dumitru Roman (Eds.). Lecture Notes in Computer Science, vol. 11784. Springer, 36–51.
Remy Delanaux, Angela Bonifati, Marie-Christine Rousset, and Romuald Thion. 2018. Query-Based
Linked Data Anonymization. In The Semantic Web - ISWC 2018 - 17th International Semantic Web
Conference, Monterey, CA, USA, October 8-12, 2018, Proceedings, Part I, Denny Vrandečić, Kalina
Bontcheva, Mari Carmen Suárez-Figueroa, Valentina Presutti, Irene Celino, Marta Sabou, Lucie-
Aimée Kaffee, and Elena Simperl (Eds.). Lecture Notes in Computer Science, vol. 11136. Springer,
530–546.
Thomas Demeester, Tim Rocktäschel, and Sebastian Riedel. 2016. Lifted Rule Injection for Relation
Embeddings. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language
Processing, EMNLP 2016, Austin, Texas, USA, November 1-4, 2016, Jian Su, Xavier Carreras, and Kevin
Duh (Eds.). The Association for Computational Linguistics, 1389–1399.
Dong Deng, Yu Jiang, Guoliang Li, Jian Li, and Cong Yu. 2013. Scalable Column Concept Determination
for Web Tables Using Large Knowledge Bases. Proceedings of the VLDB Endowment 6(13), 1606–1617.
Tim Dettmers, Pasquale Minervini, Pontus Stenetorp, and Sebastian Riedel. 2018. Convolutional 2D
Knowledge Graph Embeddings. In Proceedings of the Thirty-Second AAAI Conference on Artificial
Intelligence, (AAAI-18), the 30th innovative Applications of Artificial Intelligence (IAAI-18), and the 8th
AAAI Symposium on Educational Advances in Artificial Intelligence (EAAI-18), New Orleans, Louisiana,
USA, February 2-7, 2018, Sheila A. McIlraith and Kilian Q. Weinberger (Eds.). AAAI Press, 1811–
1818.
Deepika Devarajan. 2017. Happy Birthday Watson Discovery. IBM Cloud Blog. December 21, 2017.
Gonzalo I. Diaz, Marcelo Arenas, and Michael Benedikt. 2016. SPARQLByE: Querying RDF data by exam‐
ple. Proceedings of the VLDB Endowment 9(13), 1533–1536.
Dennis Diefenbach, Andreas Both, Kamal Singh, and Pierre Maret. 2020. Towards a question answering
system over the Semantic Web. Semantic Web Journal 11(3), 421–439.
Rose Dieng, Alain Giboin, Paul-André Tourtier, and Olivier Corby. 1992. Knowledge Acquisition for
Explainable, Multi-Expert, Knowledge-Based Design Systems. In Current Developments in Knowledge
Acquisition - EKAW'92, 6th European Knowledge Acquisition Workshop, Heidelberg and Kaiserslautern,
Germany, May 18-22, 1992, Thomas Wetter, Klaus-Dieter Althoff, John H. Boose, Brian R. Gaines,
and Marc Linster (Eds.). Lecture Notes in Computer Science, vol. 599. Springer, 298–317.
Give us feedback!
https://kgbook.org 113/150
2/15/23, 5:13 PM Knowledge Graphs
Anastasia Dimou, Miel Vander Sande, Jason Slepicka, Pedro A. Szekely, Erik Mannens, Craig A.
Knoblock, and Rik Van de Walle. 2014. Mapping Hierarchical Sources into RDF Using the RML
Mapping Language. In 2014 IEEE International Conference on Semantic Computing, Newport Beach, CA,
USA, June 16-18, 2014. IEEE Computer Society, 151–158.
Renata Queiroz Dividino, Sergej Sizov, Steffen Staab, and Bernhard Schueler. 2009. Querying for prove‐
nance, trust, uncertainty and other meta knowledge in RDF. Journal of Web Semantics 7(3), 204–219.
Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Murphy, Thomas Strohmann,
Shaohua Sun, and Wei Zhang. 2014. Knowledge vault: a web-scale approach to probabilistic knowl‐
edge fusion. In The 20th ACM SIGKDD International Conference on Knowledge Discovery and Data
Mining, KDD '14, New York, NY, USA - August 24 - 27, 2014, Sofus A. Macskassy, Claudia Perlich,
Jure Leskovec, Wei Wang, and Rayid Ghani (Eds.). ACM Press, 601–610.
Xin Luna Dong. 2019. Building a Broad Knowledge Graph for Products. In 35th IEEE International
Conference on Data Engineering, ICDE 2019, Macao, China, April 8-11, 2019. IEEE Computer Society,
25–25.
Uwe Draisbach and Felix Naumann. 2011. A generalization of blocking and windowing algorithms for
duplicate detection. In 2011 International Conference on Data and Knowledge Engineering, ICDKE 2011,
Table of Contents
Milano, Italy, September 6, 2011, Ji Zhang and Giovanni Livraga (Eds.). IEEE Computer Society, 18–
24.
Martin Dürst and Michel Suignard. 2005. Internationalized Resource Identifiers (IRIs). RFC. Internet
Engineering Task Force. January, 2005.
Arnab Dutta, Christian Meilicke, and Heiner Stuckenschmidt. 2014. Semantifying Triples from Open
Information Extraction Systems. In STAIRS 2014 - Proceedings of the 7th European Starting AI
Researcher Symposium, Prague, Czech Republic, August 18-22, 2014, Ulle Endriss and João Leite (Eds.).
Frontiers in Artificial Intelligence and Applications, vol. 264. IOS Press, 111–120.
Arnab Dutta, Christian Meilicke, and Heiner Stuckenschmidt. 2015. Enriching Structured Knowledge with
Open Information. In Proceedings of the 24th International Conference on World Wide Web, WWW
2015, Florence, Italy, May 18-22, 2015, Aldo Gangemi, Stefano Leonardi, and Alessandro Panconesi
(Eds.). ACM Press, 267–277.
Cynthia Dwork. 2006. Differential Privacy. In Automata, Languages and Programming, 33rd International
Colloquium, ICALP 2006, Venice, Italy, July 10-14, 2006, Proceedings, Part II, Michele Bugliesi, Bart
Preneel, Vladimiro Sassone, and Ingo Wegener (Eds.). Lecture Notes in Computer Science, vol. 4052.
Springer, 1–12.
Eugene Dynkin. 1965. Markov processes: Volume 1. Springer.
Julian Eberius, Katrin Braunschweig, Markus Hentsch, Maik Thiele, Ahmad Ahmadov, and Wolfgang
Lehner. 2015. Building the Dresden Web Table Corpus: A Classification Approach. In 2nd IEEE/ACM
International Symposium on Big Data Computing, BDC 2015, Limassol, Cyprus, December 7-10, 2015,
Ioan Raicu, Omer F. Rana, and Rajkumar Buyya (Eds.). IEEE Computer Society, 41–50.
Mikel Egaña, Alan Rector, Robert Stevens, and Erick Antezana. 2008. Applying Ontology Design Patterns
in Bio-ontologies. In Knowledge Engineering: Practice and Patterns, 16th International Conference, EKAW
2008, Acitrezza, Italy, September 29 - October 2, 2008. Proceedings, Aldo Gangemi and Jérôme Euzenat
(Eds.). Lecture Notes in Computer Science, vol. 5268. Springer, 7–16.
Lisa Ehrlinger and Wolfram Wöß. 2016. Towards a Definition of Knowledge Graphs. In Joint Proceedings
of the Posters and Demos Track of the 12th International Conference on Semantic Systems -
SEMANTiCS2016 and the 1st International Workshop on Semantic Change & Evolving Semantics (SuC‐
CESS'16) co-located with the 12th International Conference on Semantic Systems (SEMANTiCS 2016),
Leipzig, Germany, September 12-15, 2016, Michael Martin, Martí Cuquet, and Erwin Folmer (Eds.).
CEUR Workshop Proceedings, vol. 1695. Sun SITE Central Europe (CEUR). 4 pages.
Shady Elbassuoni, Maya Ramanath, Ralf Schenkel, Marcin Sydow, and Gerhard Weikum. 2009.
Language-model-based ranking for queries on RDF-graphs. In Proceedings of the 18th ACM Conference
on Information and Knowledge Management, CIKM 2009, Hong Kong, China, November 2-6, 2009, David
Wai-Lok Cheung, Il-Yeol Song, Wesley W. Chu, Xiaohua Hu, and Jimmy J. Lin (Eds.). ACM Press,
977–986.
Basil Ell, Andreas Harth, and Elena Simperl. 2014. SPARQL Query Verbalization for Explaining Semantic
Search Engine Queries. In The Semantic Web: Trends and Challenges - 11th International Conference,
ESWC 2014, Anissaras, Crete, Greece, May 25-29, 2014. Proceedings, Valentina Presutti, Claudia d’Am‐
Give us feedback!
https://kgbook.org 114/150
2/15/23, 5:13 PM Knowledge Graphs
ato, Fabien Gandon, Mathieu d’Aquin, Stephen Staab, and Anna Tordia (Eds.). Lecture Notes in
Computer Science, vol. 8465. Springer, 426–441.
Orri Erling. 2012. Virtuoso, a Hybrid RDBMS/Graph Column Store. IEEE Data Engineering Bulletin 35(1),
3–8.
Ivan Ermilov and Axel-Cyrille Ngonga Ngomo. 2016. TAIPAN: Automatic Property Mapping for Tabular
Data. In Knowledge Engineering and Knowledge Management - 20th International Conference, EKAW
2016, Bologna, Italy, November 19-23, 2016, Proceedings, Eva Blomqvist, Paolo Ciancarini, Francesco
Poggi, and Fabio Vitali (Eds.). Lecture Notes in Computer Science, vol. 10024. Springer, 163–179.
Diego Esteves, Anisa Rula, Aniketh Janardhan Reddy, and Jens Lehmann. 2018. Toward Veracity
Assessment in RDF Knowledge Bases: An Exploratory Analysis. Journal of Data and Information
Quality 9(3), 16:1–16:26.
Ernesto Estrada. 2011. The Structure of Complex Networks: Theory and Applications. Oxford University
Press, Inc..
Oren Etzioni, Michael J. Cafarella, Doug Downey, Stanley Kok, Ana-Maria Popescu, Tal Shaked, Stephen
Soderland, Daniel S. Weld, and Alexander Yates. 2004. Web-scale information extraction in know‐
itall: (preliminary results). In Proceedings of the 13th international conference on World Wide Web,
Table of Contents
WWW 2004, New York, NY, USA, May 17-20, 2004, Stuart I. Feldman, Mike Uretsky, Marc Najork,
and Craig E. Wills (Eds.). ACM Press, 100–110.
Oren Etzioni, Anthony Fader, Janara Christensen, Stephen Soderland, and Mausam. 2011. Open
Information Extraction: The Second Generation. In IJCAI 2011, Proceedings of the 22nd International
Joint Conference on Artificial Intelligence, Barcelona, Catalonia, Spain, July 16-22, 2011, Toby Walsh
(Ed.). IJCAI/AAAI, 3–10.
Anthony Fader, Stephen Soderland, and Oren Etzioni. 2011. Identifying Relations for Open Information
Extraction. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing,
EMNLP 2011, 27-31 July 2011, John McIntyre Conference Centre, Edinburgh, UK, A meeting of SIGDAT,
a Special Interest Group of the ACL, Regina Barzilay and Mark Johnson (Eds.). The Association for
Computational Linguistics, 1535–1545.
Wenfei Fan, Xin Wang, and Yinghui Wu. 2013. Diversified Top-$k$ Graph Pattern Matching. Proceedings
of the VLDB Endowment 6(13), 1510–1521.
Nicola Fanizzi, Claudia d’Amato, and Floriana Esposito. 2008. DL-FOIL Concept Learning in Description
Logics. In Inductive Logic Programming, 18th International Conference, ILP 2008, Prague, Czech Republic,
September 10-12, 2008, Proceedings, Filip Zelezný and Nada Lavrac (Eds.). Lecture Notes in Computer
Science, vol. 5194. Springer, 107–121.
Michael Färber, Frederic Bartscherer, Carsten Menne, and Achim Rettinger. 2018. Linked data quality of
DBpedia, Freebase, OpenCyc, Wikidata, and YAGO. Semantic Web Journal 9(1), 77–129.
Michael Färber. 2019. The Microsoft Academic Knowledge Graph: A Linked Data Source with 8 Billion
Triples of Scholarly Data. In The Semantic Web - ISWC 2019 - 18th International Semantic Web
Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part II, Chiara Ghidini, Olaf
Hartig, Maria Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime
Lefrançois, and Fabien Gandon (Eds.). Lecture Notes in Computer Science, vol. 11779. Springer,
113–129.
Dieter Fensel, Umutcan Simsek, Kevin Angele, Elwin Huaman, Elias Kärle, Oleksandra Panasiuk, Ioan
Toma, Jürgen Umbrich, and Alexander Wahler. 2020. Knowledge Graphs - Methodology, Tools and
Selected Use Cases. Springer.
Mariano Fernández, Asunción Gómez-Pérez, and Natalia Juristo. 1997. METHONTOLOGY: from
Ontological Art towards Ontological Engineering. In Ontological Engineering, Papers from the AAAI
Spring Symposium, Technical Report SS-97-06, 1997, Adam Farquhar and Michael Gruninger (Eds.).
AAAI Press, 33–40.
Javier D. Fernández, Miguel A. Martínez-Prieto, Claudio Gutiérrez, Axel Polleres, and Mario Arias. 2013.
Binary RDF representation for publication and exchange (HDT). Journal of Web Semantics 19, 22–41.
Javier D. Fernández, Sabrina Kirrane, Axel Polleres, and Simon Steyskal. 2017. Self-Enforcing Access
Control for Encrypted RDF. In The Semantic Web - 14th International Conference, ESWC 2017, Portorož,
Slovenia, May 28 - June 1, 2017, Proceedings, Part I, Eva Blomqvist, Diana Maynard, Aldo Gangemi,
Rinke Hoekstra, Pascal Hitzler, and Olaf Hartig (Eds.). Lecture Notes in Computer Science,
vol. 10249. Springer, 607–622.
Give us feedback!
https://kgbook.org 115/150
2/15/23, 5:13 PM Knowledge Graphs
Emilio Ferrara, Pasquale De Meo, Giacomo Fiumara, and Robert Baumgartner. 2014. Web data extrac‐
tion, applications and techniques: A survey. Knowledge-based Systems 70, 301–323.
Sébastien Ferré. 2017. Sparklis: An expressive query builder for SPARQL endpoints with guidance in nat‐
ural language. Semantic Web Journal 8(3), 405–418.
Charles J. Fillmore. 1976. Frame semantics and the nature of language. Annals of the New York Academy
of Sciences 280(1), 20–32.
Jenny Rose Finkel, Trond Grenager, and Christopher D. Manning. 2005. Incorporating Non-local
Information into Information Extraction Systems by Gibbs Sampling. In ACL 2005, 43rd Annual
Meeting of the Association for Computational Linguistics, Proceedings of the Conference, 25-30 June 2005,
University of Michigan, USA, Kevin Knight, Hwee Tou Ng, and Kemal Oflazer (Eds.). The Association
for Computational Linguistics, 363–370.
Sergio Flesca, Giuseppe Manco, Elio Masciari, Eugenio Rende, and Andrea Tagarelli. 2004. Web wrapper
induction: a brief survey. AI Communications 17(2), 57–61.
Giorgos Flouris, Irini Fundulaki, Maria Michou, and Grigoris Antoniou. 2010. Controlling Access to RDF
Graphs. In Future Internet - FIS 2010 - Third Future Internet Symposium, Berlin, Germany, September 20-
22, 2010. Proceedings, Arne-Jørgen Berre, Asunción Gómez-Pérez, Kurt Tutschku, and Dieter Fensel
Table of Contents
https://kgbook.org 116/150
2/15/23, 5:13 PM Knowledge Graphs
Aldo Gangemi. 2005. Ontology design patterns for semantic web content. In The Semantic Web - ISWC
2005, 4th International Semantic Web Conference, ISWC 2005, Galway, Ireland, November 6-10, 2005,
Proceedings, Yolanda Gil, Enrico Motta, V. Richard Benjamins, and Mark A. Musen (Eds.). Lecture
Notes in Computer Science, vol. 3729. Springer, 262–276.
Claire Gardent, Anastasia Shimorina, Shashi Narayan, and Laura Perez-Beltrachini. 2017. The WebNLG
challenge: Generating text from RDF data. In Proceedings of the 10th International Conference on
Natural Language Generation, INLG 2017, Santiago de Compostela, Spain, September 4-7, 2017, José M.
Alonso, Alberto Bugarín, and Ehud Reiter (Eds.). The Association for Computational Linguistics,
124–133.
Michael Gelfond and Vladimir Lifschitz. 1988. The Stable Model Semantics for Logic Programming. In
Logic Programming, Proceedings of the Fifth International Conference and Symposium, Seattle, Washington,
USA, August 15-19, 1988 (2 Volumes), Robert A. Kowalski and Kenneth A. Bowen (Eds.). The MIT
Press, 1070–1080.
James Geller, Soon Ae Chun, and Yoo Jung An. 2008. Toward the Semantic Deep Web. IEEE Computer
41(9), 95–97.
Anna Lisa Gentile, Ziqi Zhang, and Fabio Ciravegna. 2014. Self Training Wrapper Induction with Linked
Table of Contents
Data. In Text, Speech and Dialogue - 17th International Conference, TSD 2014, Brno, Czech Republic,
September 8-12, 2014. Proceedings, Petr Sojka, Ales Horák, Ivan Kopecek, and Karel Pala (Eds.).
Lecture Notes in Computer Science, vol. 8655. Springer, 285–292.
Anna Lisa Gentile, Daniel Gruhl, Petar Ristoski, and Steve Welch. 2019. Personalized Knowledge Graphs
for the Pharmaceutical Domain. In The Semantic Web - ISWC 2019 - 18th International Semantic Web
Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part II, Chiara Ghidini, Olaf
Hartig, Maria Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime
Lefrançois, and Fabien Gandon (Eds.). Lecture Notes in Computer Science, vol. 11779. Springer,
400–417.
Daniel Gerber, Diego Esteves, Jens Lehmann, Lorenz Bühmann, Ricardo Usbeck, Axel-Cyrille Ngonga
Ngomo, and René Speck. 2015. DeFacto–temporal and multilingual deep fact validation. Journal of
Web Semantics 35, 85–101.
Sabrina Gerbracht. 2008. Possibilities to Encrypt an RDF-Graph. In 2008 3rd International Conference on
Information and Communication Technologies: From Theory to Applications. IEEE Computer Society. 6
pages.
Lise Getoor and Ben Taskar (Eds.). 2007. Introduction to Statistical Relational Learning. Adaptive
Computation and Machine Learning. The MIT Press.
Mark Giereth. 2005. On Partial Encryption of RDF-Graphs. In The Semantic Web - ISWC 2005, 4th
International Semantic Web Conference, ISWC 2005, Galway, Ireland, November 6-10, 2005, Proceedings,
Yolanda Gil, Enrico Motta, V. Richard Benjamins, and Mark A. Musen (Eds.). Lecture Notes in
Computer Science, vol. 3729. Springer, 308–322.
Yolanda Gil, Simon Miles, Khalid Belhajjame, Daniel Garijo, Graham Klyne, Paolo Missier, Stian Soiland-
Reyes, and Stephan Zednik. 2013. PROV Model Primer. W3C Working Group Note. World Wide Web
Consortium. April 30, 2013.
José M. Giménez-García, Antoine Zimmermann, and Pierre Maret. 2017. NdFluents: An Ontology for
Annotated Statements with Inference Preservation. In The Semantic Web - 14th International
Conference, ESWC 2017, Portorož, Slovenia, May 28 - June 1, 2017, Proceedings, Part I, Eva Blomqvist,
Diana Maynard, Aldo Gangemi, Rinke Hoekstra, Pascal Hitzler, and Olaf Hartig (Eds.). Lecture Notes
in Computer Science, vol. 10249. Springer, 638–654.
Birte Glimm. 2011. Using SPARQL with RDFS and OWL Entailment. In Reasoning Web. Semantic
Technologies for the Web of Data - 7th International Summer School 2011, Galway, Ireland, August 23-27,
2011, Tutorial Lectures, Axel Polleres, Claudia d’Amato, Marcelo Arenas, Siegfried Handschuh, Paula
Kroner, Sascha Ossowski, and Peter F. Patel-Schneider (Eds.). Lecture Notes in Computer Science,
vol. 6848. Springer, 137–201.
Xavier Glorot, Antoine Bordes, Jason Weston, and Yoshua Bengio. 2013. A Semantic Matching Energy
Function for Learning with Multi-relational Data. In 1st International Conference on Learning
Representations, ICLR 2013, Scottsdale, Arizona, USA, May 2-4, 2013, Workshop Track Proceedings,
Yoshua Bengio and Yann LeCun (Eds.). OpenReview.net. 4 pages.
Patrick Golden and Ryan B. Shaw. 2016. Nanopublication beyond the sciences: the PeriodO period
gazetteer. PeerJ Computer Science 2, pe44. Give us feedback!
https://kgbook.org 117/150
2/15/23, 5:13 PM Knowledge Graphs
Asunción Gómez-Pérez, Mariano Fernández-López, and Oscar Corcho. 2006. Ontological Engineering: with
examples from the areas of Knowledge Management, e-Commerce and the Semantic Web. Advanced
Information and Knowledge Processing. Springer.
Rafael S. Gonçalves, Matthew Horridge, Rui Li, Yu Liu, Mark A. Musen, Csongor I. Nyulas, Evelyn
Obamos, Dhananjay Shrouty, and David Temple. 2019. Use of OWL and Semantic Web Technologies
at Pinterest. In The Semantic Web - ISWC 2019 - 18th International Semantic Web Conference, Auckland,
New Zealand, October 26-30, 2019, Proceedings, Part II, Chiara Ghidini, Olaf Hartig, Maria
Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime Lefrançois, and Fabien
Gandon (Eds.). Lecture Notes in Computer Science, vol. 11779. Springer, 418–435.
Larry González and Aidan Hogan. 2018. Modelling Dynamics in Semantic Web Knowledge Graphs with
Formal Concept Analysis. In Proceedings of the 2018 World Wide Web Conference on World Wide Web,
WWW 2018, Lyon, France, April 23-27, 2018, Pierre-Antoine Champin, Fabien L. Gandon, Mounia
Lalmas, and Panagiotis G. Ipeirotis (Eds.). ACM Press, 1175–1184.
Simon Gottschalk and Elena Demidova. 2018. EventKG: A Multilingual Event-Centric Temporal
Knowledge Graph. In The Semantic Web - 15th International Conference, ESWC 2018, Heraklion, Crete,
Greece, June 3-7, 2018, Proceedings, Aldo Gangemi, Roberto Navigli, Maria-Esther Vidal, Pascal
Table of Contents
Hitzler, Troncy Raphaël, Laura Hollink, Anna Tordai, and Mehwish Alam (Eds.). Lecture Notes in
Computer Science, vol. 10843. Springer, 272–287.
Guido Governatori, Ho-Pun Lam, Antonino Rotolo, Serena Villata, and Fabien Gandon. 2013. Heuristics
for Licenses Composition. In Legal Knowledge and Information Systems - JURIX 2013: The Twenty-Sixth
Annual Conference, December 11-13, 2013, University of Bologna, Italy, Kevin D. Ashley (Ed.). Frontiers
in Artificial Intelligence and Applications, vol. 259. IOS Press, 77–86.
Ralph Grishman. 2012. Information Extraction: Capabilities and Challenges. Notes prepared for the 2012
International Winter School in Language and Speech Technologies.
Paul T. Groth, Antonis Loizou, Alasdair J. G. Gray, Carole A. Goble, Lee Harland, and Steve Pettifer.
2014. API-centric Linked Data integration: The Open PHACTS Discovery Platform case study. Journal
of Web Semantics 29, 12–18.
Michael Grüninger and Mark S. Fox. 1995. Methodology for the Design and Evaluation of Ontologies. In
Workshop on Basic Ontological Issues in Knowledge Sharing, IJCAI-95, Montreal. 10 pages.
Michael Grüninger and Mark S. Fox. 1995. The role of competency questions in enterprise engineering. In
Benchmarking – Theory and practice, (Ed.). Springer, 22–31.
Nicola Guarino, Daniel Oberle, and Steffen Staab. 2009. What Is an Ontology?. In Handbook on Ontologies,
Steffen Staab and Rudi Studer (Eds.). International Handbooks on Information Systems. Springer, 1–
17.
Ramanathan V. Guha, Rob McCool, and Eric Miller. 2003. Semantic search. In Proceedings of the Twelfth
International World Wide Web Conference, WWW2003, Budapest, Hungary, 20-24 May 2003, Gusztáv
Hencsey, Bebo White, Yih-Farn Robin Chen, László Kovács, and Steve Lawrence (Eds.). ACM Press,
700–709.
Ramanathan V. Guha, Rob McCool, and Richard Fikes. 2004. Contexts for the Semantic Web. In The
Semantic Web - ISWC 2004: Third International Semantic Web Conference, Hiroshima, Japan, November
7-11, 2004. Proceedings, Frank van Harmelen, Sheila McIlraith, and Dimitri Plexousakis (Eds.).
Lecture Notes in Computer Science, vol. 3298. Springer, 32–46.
Shu Guo, Quan Wang, Lihong Wang, Bin Wang, and Li Guo. 2016. Jointly Embedding Knowledge Graphs
and Logical Rules. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language
Processing, EMNLP 2016, Austin, Texas, USA, November 1-4, 2016, Jian Su, Xavier Carreras, and Kevin
Duh (Eds.). The Association for Computational Linguistics, 192–202.
Shu Guo, Quan Wang, Lihong Wang, Bin Wang, and Li Guo. 2018. Knowledge Graph Embedding With
Iterative Guidance From Soft Rules. In Proceedings of the Thirty-Second AAAI Conference on Artificial
Intelligence, (AAAI-18), the 30th innovative Applications of Artificial Intelligence (IAAI-18), and the 8th
AAAI Symposium on Educational Advances in Artificial Intelligence (EAAI-18), New Orleans, Louisiana,
USA, February 2-7, 2018, Sheila A. McIlraith and Kilian Q. Weinberger (Eds.). AAAI Press, 4816–
4823.
Sonal Gupta and Christopher D. Manning. 2014. Improved Pattern Learning for Bootstrapped Entity
Extraction. In Proceedings of the Eighteenth Conference on Computational Natural Language Learning,
CoNLL 2014, Baltimore, Maryland, USA, June 26-27, 2014, Roser Morante and Wen-tau Yih (Eds.).
The Association for Computational Linguistics, 98–108. Give us feedback!
https://kgbook.org 118/150
2/15/23, 5:13 PM Knowledge Graphs
Claudio Gutiérrez, Carlos A. Hurtado, and Alejandro A. Vaisman. 2007. Introducing Time into RDF. IEEE
Transactions on Knowledge and Data Engineering 19(2), 207–218.
Florian Haag, Steffen Lohmann, Stephan Siek, and Thomas Ertl. 2015. QueryVOWL: A Visual Query
Notation for Linked Data. In The Semantic Web: ESWC 2015 Satellite Events - ESWC 2015 Satellite
Events Portorož, Slovenia, May 31 - June 4, 2015, Revised Selected Papers, Fabien Gandon, Christophe
Guéret, Serena Villata, John G. Breslin, Catherine Faron-Zucker, and Antoine Zimmermann (Eds.).
Lecture Notes in Computer Science, vol. 9341. Springer, 387–402.
Juha Hakala. 2010. Persistent identifiers – an overview. PersID Technical Report. September 3, 2010.
Ferras Hamad, Isaac Liu, and Xian Xing Zhang. 2018. Food Discovery with Uber Eats: Building a Query
Understanding Engine. Uber Engineering Blog. june 10, 2018.
William L. Hamilton, Payal Bajaj, Marinka Zitnik, Dan Jurafsky, and Jure Leskovec. 2018. Embedding
Logical Queries on Knowledge Graphs. In Advances in Neural Information Processing Systems 31:
Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, 3-8 December 2018,
Montréal, Canada, Samy Bengio, Hanna M. Wallach, Hugo Larochelle, Kristen Grauman, Nicolò Cesa-
Bianchi, and Roman Garnett (Eds.), 2030–2041.
Tony Hammond, Michele Pasin, and Evangelos Theodoridis. 2017. Data integration and disintegration:
Table of Contents
Managing Springer Nature SciGraph with SHACL and OWL. In Proceedings of the ISWC 2017 Posters &
Demonstrations and Industry Tracks co-located with 16th International Semantic Web Conference (ISWC
2017), Vienna, Austria, October 23rd - to - 25th, 2017, Nadeschda Nikitina, Dezhao Song, Achille
Fokoue, and Peter Haase (Eds.). CEUR Workshop Proceedings, vol. 1963. Sun SITE Central Europe
(CEUR). 2 pages.
Steve Harris, Andy Seaborne, and Eric Prud’hommeaux. 2013. SPARQL 1.1 Query Language. W3C
Recommendation. World Wide Web Consortium. March 21, 2013.
Andreas Harth. 2010. VisiNav: A system for visual search and navigation on web data. Journal of Web
Semantics 8(4), 348–354.
Olaf Hartig and Carlos Buil-Aranda. 2016. Bindings-Restricted Triple Pattern Fragments. In On the Move
to Meaningful Internet Systems: OTM 2016 Conferences - Confederated International Conferences: CoopIS,
C&TC, and ODBASE 2016, Rhodes, Greece, October 24-28, 2016, Proceedings, Christophe Debruyne,
Hervé Panetto, Robert Meersman, Tharam S. Dillon, eva Kühn, Declan O’Sullivan, and Claudio
Agostino Ardagna (Eds.). Lecture Notes in Computer Science, vol. 10033. Springer, 762–779.
Olaf Hartig and Bryan Thompson. 2014. Foundations of an Alternative Approach to Reification in RDF.
CoRR abs/1406.3399. 14 pages.
Olaf Hartig, Christian Bizer, and Johann Christoph Freytag. 2009. Executing SPARQL Queries over the
Web of Linked Data. In The Semantic Web - ISWC 2009, 8th International Semantic Web Conference,
ISWC 2009, Chantilly, VA, USA, October 25-29, 2009. Proceedings, Abraham Bernstein, David R.
Karger, Tom Heath, Lee Feigenbaum, Diana Maynard, Enrico Motta, and Krishnaprasad
Thirunarayan (Eds.). Lecture Notes in Computer Science, vol. 5823. Springer, 293–309.
Olaf Hartig, Ian Letter, and Jorge Pérez. 2017. A Formal Framework for Comparing Linked Data
Fragments. In The Semantic Web - ISWC 2017 - 16th International Semantic Web Conference, Vienna,
Austria, October 21-25, 2017, Proceedings, Part I, Claudia d’Amato, Miriam Fernández, Valentina A.
M. Tamma, Freddy Lécué, Philippe Cudré-Mauroux, Juan F. Sequeda, Christoph Lange, and Jeff
Heflin (Eds.). Lecture Notes in Computer Science, vol. 10587. Springer, 364–382.
Olaf Hartig. 2017. Foundations of RDF* and SPARQL* – An Alternative Approach to Statement-Level
Metadata in RDF. In Proceedings of the 11th Alberto Mendelzon International Workshop on Foundations
of Data Management and the Web, Montevideo, Uruguay, June 7-9, 2017, Juan L. Reutter and Divesh
Srivastava (Eds.). CEUR Workshop Proceedings, vol. 1912. Sun SITE Central Europe (CEUR). 11
pages.
Patrick J. Hayes and Peter F. Patel-Schneider. 2014. RDF 1.1 Semantics. W3C Recommendation. World
Wide Web Consortium. February 25, 2014.
Qi He, Bee-Chung Chen, and Deepak Agarwal. 2016. Building The LinkedIn Knowledge Graph. LinkedIn
Blog. October 6, 2016.
Marti A. Hearst. 1992. Automatic Acquisition of Hyponyms from Large Text Corpora. In 14th International
Conference on Computational Linguistics, COLING 1992, Nantes, France, August 23-28, 1992, 539–545.
Tom Heath and Christian Bizer. 2011. Linked Data: Evolving the Web into a Global Data Space (1st Edition).
Synthesis Lectures on the Semantic Web: Theory and Technology, vol. 1. Morgan & Claypool.
Give us feedback!
https://kgbook.org 119/150
2/15/23, 5:13 PM Knowledge Graphs
Tom Heath and Enrico Motta. 2008. Revyu: Linking reviews and ratings into the Web of Data. Journal of
Web Semantics 6(4), 266–273.
Stefan Heindorf, Martin Potthast, Benno Stein, and Gregor Engels. 2016. Vandalism Detection in
Wikidata. In Proceedings of the 25th ACM International Conference on Information and Knowledge
Management, CIKM 2016, Indianapolis, IN, USA, October 24-28, 2016, Snehasis Mukhopadhyay,
ChengXiang Zhai, Elisa Bertino, Fabio Crestani, Javed Mostafa, Jie Tang, Luo Si, Xiaofang Zhou, Yi
Chang, Yunyao Li, and Parikshit Sondhi (Eds.). ACM Press, 327–336.
Benjamin Heitmann, Felix Hermsen, and Stefan Decker. 2017. k-RDF-Neighbourhood Anonymity:
Combining Structural and Attribute-based Anonymisation for Linked Data. In Proceedings of the 5th
Workshop on Society, Privacy and the Semantic Web - Policy and Technology (PrivOn2017) co-located
with 16th International Semantic Web Conference (ISWC 2017), Vienna, Austria, October 22, 2017,
Christopher Brewster, Michelle Cheatham, Mathieu d’Aquin, Stefan Decker, and Sabrina Kirrane
(Eds.). CEUR Workshop Proceedings, vol. 1951. Sun SITE Central Europe (CEUR). 16 pages.
Sebastian Hellmann, Jens Lehmann, Sören Auer, and Martin Brümmer. 2013. Integrating NLP Using
Linked Data. In The Semantic Web - ISWC 2013 - 12th International Semantic Web Conference, Sydney,
NSW, Australia, October 21-25, 2013, Proceedings, Part II, Harith Alani, Lalana Kagal, Achille Fokoue,
Table of Contents
Paul Groth, Chris Biemann, Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty, and
Krzysztof Janowicz (Eds.). Lecture Notes in Computer Science, vol. 8219. Springer, 98–113.
Remko Helms and Kees Buijsrogge. 2005. Knowledge Network Analysis: A Technique to Analyze
Knowledge Management Bottlenecks in Organizations. In 16th International Workshop on Database
and Expert Systems Applications (DEXA 2005), 22-26 August 2005, Copenhagen, Denmark. IEEE
Computer Society, 410–414.
James A. Hendler, Jeanne Holm, Chris Musialek, and George Thomas. 2012. US Government Linked
Open Data: Semantic.data.gov. IEEE Intelligent Systems 27(3), 25–31.
Cory Henson, Stefan Schmid, Anh Tuan Tran, and Antonios Karatzoglou. 2019. Using a Knowledge Graph
of Scenes to Enable Search of Autonomous Driving Data. In Proceedings of the ISWC 2019 Satellite
Tracks (Posters & Demonstrations, Industry, and Outrageous Ideas) co-located with 18th International
Semantic Web Conference (ISWC 2019), Auckland, New Zealand, October 26-30, 2019, Mari Carmen
Suárez-Figueroa, Gong Cheng, Anna Lisa Gentile, Christophe Guéret, C. Maria Keet, and Abraham
Bernstein (Eds.). CEUR Workshop Proceedings, vol. 2456. Sun SITE Central Europe (CEUR), 313–
314.
Daniel Hernández, Aidan Hogan, and Markus Krötzsch. 2015. Reifying RDF: What Works Well With
Wikidata?. In Proceedings of the 11th International Workshop on Scalable Semantic Web Knowledge Base
Systems co-located with 14th International Semantic Web Conference (ISWC 2015), Bethlehem, PA, USA,
October 11, 2015, Thorsten Liebig and Achille Fokoue (Eds.). CEUR Workshop Proceedings,
vol. 1457. Sun SITE Central Europe (CEUR), 32–47.
Frank L. Hitchcock. 1927. The Expression of a Tensor or a Polyadic as a Sum of Products. Journal of
Mathematics and Physics 6(1–4), 164–189.
Pascal Hitzler and Adila Krisnadhi. 2018. A Tutorial on Modular Ontology Modeling with Ontology
Design Patterns: The Cooking Recipes Ontology. CoRR abs/1808.08433. 22 pages.
Pascal Hitzler, Markus Krötzsch, and Sebastian Rudolph. 2010. Foundations of Semantic Web Technologies.
Chapman and Hall / CRC Press.
Pascal Hitzler, Markus Krötzsch, Bijan Parsia, Peter F. Patel-Schneider, and Sebastian Rudolph. 2012.
OWL 2 Web Ontology Language Primer (Second Edition). W3C Recommendation. World Wide Web
Consortium. December 11, 2012.
Vinh Thinh Ho, Daria Stepanova, Mohamed H. Gad-Elrab, Evgeny Kharlamov, and Gerhard Weikum.
2018. Rule Learning from Knowledge Graphs Guided by Embedding Models. In The Semantic Web -
ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA, USA, October 8-12, 2018,
Proceedings, Part I, Denny Vrandečić, Kalina Bontcheva, Mari Carmen Suárez-Figueroa, Valentina
Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture Notes in
Computer Science, vol. 11136. Springer, 72–90.
Cornelis Hoede. 1995. On the ontology of knowledge graphs. In Conceptual Structures: Applications,
Implementation and Theory, Third International Conference on Conceptual Structures, ICCS '95, Santa
Cruz, California, USA, August 14-18, 1995, Proceedings, Gerard Ellis, Robert Levinson, William Rich,
and John F. Sowa (Eds.). Lecture Notes in Computer Science, vol. 954. Springer, 308–322.
Give us feedback!
https://kgbook.org 120/150
2/15/23, 5:13 PM Knowledge Graphs
Johannes Hoffart, Fabian M. Suchanek, Klaus Berberich, Edwin Lewis-Kelham, Gerard de Melo, and
Gerhard Weikum. 2011. YAGO2: Exploring and Querying World Knowledge in Time, Space, Context,
and Many Languages. In Proceedings of the 20th International Conference on World Wide Web, WWW
2011, Hyderabad, India, March 28 - April 1, 2011 (Companion Volume), Sadagopan Srinivasan, Krithi
Ramamritham, Arun Kumar, M. P. Ravindra, Elisa Bertino, and Ravi Kumar (Eds.). ACM Press, 229–
232.
Raphael Hoffmann, Congle Zhang, Xiao Ling, Luke S. Zettlemoyer, and Daniel S. Weld. 2011. Knowledge-
Based Weak Supervision for Information Extraction of Overlapping Relations. In The 49th Annual
Meeting of the Association for Computational Linguistics: Human Language Technologies, Proceedings of
the Conference, 19-24 June, 2011, Portland, Oregon, USA, Dekang Lin, Yuji Matsumoto, and Rada
Mihalcea (Eds.). The Association for Computational Linguistics, 541–550.
Aidan Hogan. 2017. Canonical Forms for Isomorphic and Equivalent RDF Graphs: Algorithms for Leaning
and Labelling Blank Nodes. ACM Transactions on the Web 11(4), 22:1–22:62.
Aidan Hogan. 2020. Knowledge Graphs: Research Directions. In Reasoning Web. Declarative Artificial
Intelligence – 16th International Summer School 2020, Oslo, Norway, June 24–26, 2020, Tutorial
Lectures, Marco Manna and Andreas Pieris (Eds.). Lecture Notes in Computer Science, vol. 12258.
Table of Contents
Springer, 223–253.
Aidan Hogan. 2020. The Web of Data. Springer.
Aidan Hogan, Andreas Harth, Alexandre Passant, Stefan Decker, and Axel Polleres. 2010. Weaving the
Pedantic Web. In Proceedings of the WWW2010 Workshop on Linked Data on the Web, LDOW 2010,
Raleigh, USA, April 27, 2010, Christian Bizer, Tom Heath, Tim Berners-Lee, and Michael Hausenblas
(Eds.). CEUR Workshop Proceedings, vol. 628. Sun SITE Central Europe (CEUR). 10 pages.
Aidan Hogan, Jürgen Umbrich, Andreas Harth, Richard Cyganiak, Axel Polleres, and Stefan Decker.
2012. An empirical survey of Linked Data conformance. Journal of Web Semantics 14, 14–44.
Aidan Hogan, Antoine Zimmermann, Jürgen Umbrich, Axel Polleres, and Stefan Decker. 2012. Scalable
and distributed methods for entity matching, consolidation and disambiguation over Linked Data
corpora. Journal of Web Semantics 10, 76–110.
Aidan Hogan, Marcelo Arenas, Alejandro Mallea, and Axel Polleres. 2014. Everything you always wanted
to know about blank nodes. Journal of Web Semantics 27–28, 42–69.
Aidan Hogan, Juan L. Reutter, and Adrián Soto. 2020. In-Database Graph Analytics with Recursive
SPARQL. In The Semantic Web - ISWC 2020 - 19th International Semantic Web Conference, Athens,
Greece, November 2-6, 2020, Proceedings, Part I, Jeff Z. Pan, Valentina A. M. Tamma, Claudia d’Ama‐
to, Krzysztof Janowicz, Bo Fu, Axel Polleres, Oshani Seneviratne, and Lalana Kagal (Eds.). Lecture
Notes in Computer Science, vol. 12506. Springer, 511–528.
Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia d’Amato, Gerard de Melo, Claudio Gutiérrez,
Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier, Axel-Cyrille Ngonga
Ngomo, Axel Polleres, Sabbir M. Rashid, Anisa Rula, Lukas Schmelzeisen, Juan F. Sequeda, Steffen
Staab, and Antoine Zimmermann. 2021. Knowledge Graphs. ACM Computing Surveys 54(4), 71:1–
71:37.
Ian Horrocks and Peter F. Patel-Schneider. 2004. Reducing OWL entailment to description logic satisfia‐
bility. Journal of Web Semantics 1(4), 345–357.
Ian Horrocks, Peter F. Patel-Schneider, Harold Boley, Said Tabet, Benjamin Grosof, and Mike Dean. 2004.
SWRL: A Semantic Web Rule Language Combining OWL and RuleML. W3C Member Submission.
May 21, 2004.
Sen Hu, Lei Zou, Jeffrey Xu Yu, Haixun Wang, and Dongyan Zhao. 2018. Answering Natural Language
Questions by Subgraph Matching over Knowledge Graphs. IEEE Transactions on Knowledge and Data
Engineering 30(5), 824–837.
Xiao Huang, Jingyuan Zhang, Dingcheng Li, and Ping Li. 2019. Knowledge Graph Embedding Based
Question Answering. In Proceedings of the Twelfth ACM International Conference on Web Search and
Data Mining, WSDM 2019, Melbourne, VIC, Australia, February 11-15, 2019, J. Shane Culpepper,
Alistair Moffat, Paul N. Bennett, and Kristina Lerman (Eds.). ACM Press, 105–113.
Andy Hunt and Dave Thomas. 2003. The Trip-Packing Dilemma. IEEE Software 20(3), 106–107.
Rana Hussein, Dingqi Yang, and Philippe Cudré-Mauroux. 2018. Are Meta-Paths Necessary?: Revisiting
Heterogeneous Graph Embeddings. In Proceedings of the 27th ACM International Conference on
Information and Knowledge Management, CIKM 2018, Torino, Italy, October 22-26, 2018, Alfredo
Cuzzocrea, James Allan, Norman W. Paton, Divesh Srivastava, Rakesh Agrawal, Andrei Z. Broder,
Give us feedback!
https://kgbook.org 121/150
2/15/23, 5:13 PM Knowledge Graphs
Mohammed J. Zaki, K. Selçuk Candan, Alexandros Labrinidis, Assaf Schuster, and Haixun Wang
(Eds.). ACM Press, 437–446.
Dylan Hutchison, Bill Howe, and Dan Suciu. 2017. LaraDB: A Minimalist Kernel for Linear and Relational
Algebra Computation. In Proceedings of the 4th ACM SIGMOD Workshop on Algorithms and Systems for
MapReduce and Beyond, BeyondMR@SIGMOD 2017, Chicago, IL, USA, May 19, 2017, Foto N. Afrati
and Jacek Sroka (Eds.). ACM Press, 2:1–2:10.
Eero Hyvönen, Eetu Mäkelä, Tomi Kauppinen, Olli Alm, Jussi Kurki, Tuukka Ruotsalo, Katri Seppälä,
Joeli Takala, Kimmo Puputti, Heini Kuittinen, Kim Viljanen, Jouni Tuominen, Tuomas Palonen,
Matias Frosterus, Reetta Sinkkilä, Panu Paakkarinen, Joonas Laitio, and Katariina Nyberg. 2009.
CultureSampo: A National Publication System of Cultural Heritage on the Semantic Web 2.0. In The
Semantic Web: Research and Applications, 6th European Semantic Web Conference, ESWC 2009,
Heraklion, Crete, Greece, May 31-June 4, 2009, Proceedings, Lora Aroyo, Paolo Traverso, Fabio
Ciravegna, Philipp Cimiano, Tom Heath, Eero Hyvönen, Riichiro Mizoguchi, Eyal Oren, Marta
Sabou, and Elena Paslaru Bontas Simperl (Eds.), vol. 5554. Springer, 851–856.
Andreea Iana, Steffen Jung, Philipp Naeser, Aliaksandr Birukou, Sven Hertling, and Heiko Paulheim.
2019. Building a Conference Recommender System Based on SciGraph and WikiCFP. In Semantic
Table of Contents
Systems. The Power of AI and Knowledge Graphs - 15th International Conference, SEMANTiCS 2019,
Karlsruhe, Germany, September 9-12, 2019, Proceedings, Maribel Acosta, Philippe Cudré-Mauroux,
Maria Maleshkova, Tassilo Pellegrini, Harald Sack, and York Sure-Vetter (Eds.). Lecture Notes in
Computer Science, vol. 11702. Springer, 117–123.
Renato Iannella and Serena Villata. 2018. ODRL Information Model 2.2. W3C Recommendation. World
Wide Web Consortium. February 15, 2018.
International Health Terminology Standards Development Organisation. 2019. SNOMED CT Editorial
Guide. July 31, 2019.
Alexandru Iosup, Tim Hegeman, Wing Lung Ngai, Stijn Heldens, Arnau Prat-Pérez, Thomas Manhardt,
Hassan Chafi, Mihai Capota, Narayanan Sundaram, Michael J. Anderson, Ilie Gabriel Tanase,
Yinglong Xia, Lifeng Nai, and Peter A. Boncz. 2016. LDBC Graphalytics: A Benchmark for Large-Scale
Graph on Parallel and Distributed Platforms. Proceedings of the VLDB Endowment 9(13), 1317–1328.
Robert Isele, Anja Jentzsch, and Christian Bizer. 2011. Efficient multidimensional blocking for link dis‐
covery without losing recall. In Proceedings of the 14th International Workshop on the Web and
Databases 2011, WebDB 2011, Athens, Greece, June 12, 2011, Amélie Marian and Vasilis Vassalos
(Eds.). 6 pages.
P. James. 1992. Knowledge Graphs. In Linguistic Instruments in Knowledge Engineering, Reind P. van de Riet
and Robert A. Meersman (Eds.). Elsevier, 97–117.
Daniel Janke and Steffen Staab. 2018. Storing and Querying Semantic Data in the Cloud. In Reasoning
Web. Learning, Uncertainty, Streaming, and Scalability - 14th International Summer School 2018, Esch-
sur-Alzette, Luxembourg, September 22-26, 2018, Tutorial Lectures, Claudia d’Amato and Martin
Theobald (Eds.). Lecture Notes in Computer Science, vol. 11078. Springer, 173–222.
Daniel Janke, Steffen Staab, and Matthias Thimm. 2018. Impact analysis of data placement strategies on
query efforts in distributed RDF stores. Journal of Web Semantics 50, 21–48.
Krzysztof Janowicz, Pascal Hitzler, Benjamin Adams, Dave Kolas, and Charles Vardeman. 2014. Five stars
of Linked Data vocabulary use. Semantic Web Journal 5(3), 173–176.
Krzysztof Janowicz, Bo Yan, Blake Regalia, Rui Zhu, and Gengchen Mai. 2018. Debiasing Knowledge
Graphs: Why Female Presidents are not like Female Popes. In Proceedings of the ISWC 2018 Posters &
Demonstrations, Industry and Blue Sky Ideas Tracks co-located with 17th International Semantic Web
Conference (ISWC 2018), Monterey, USA, October 8th - to - 12th, 2018, Marieke van Erp, Medha Atre,
Vanessa López, Kavitha Srinivas, and Carolina Fortuna (Eds.). CEUR Workshop Proceedings,
vol. 2180. Sun SITE Central Europe (CEUR). 5 pages.
Nandish Jayaram, Sidharth Goyal, and Chengkai Li. 2015. VIIQ: Auto-Suggestion Enabled Visual
Interface for Interactive Graph Query Formulation. Proceedings of the VLDB Endowment 8(12), 1940–
1943.
Nandish Jayaram, Arijit Khan, Chengkai Li, Xifeng Yan, and Ramez Elmasri. 2015. Querying Knowledge
Graphs by Example Entity Tuples. IEEE Transactions on Knowledge and Data Engineering 27(10), 2797–
2811.
Guoliang Ji, Shizhu He, Liheng Xu, Kang Liu, and Jun Zhao. 2015. Knowledge graph embedding via dy‐
namic mapping matrix. In Proceedings of the 53rd Annual Meeting of the Association for Computational
Give us feedback!
https://kgbook.org 122/150
2/15/23, 5:13 PM Knowledge Graphs
Linguistics and the 7th International Joint Conference on Natural Language Processing of the Asian
Federation of Natural Language Processing, ACL 2015, July 26-31, 2015, Beijing, China, Volume 1: Long
Papers. The Association for Computational Linguistics, 687–696.
Yun-fei Jiang and Ning Ma. 2002. A Plan Recognition Algorithm Based on Plan Knowledge Graph.
Journal of Software 13(4), 686–692.
Dan Jurafsky and James H. Martin. 2021. Speech and Language Processing.
David Jurgens and Roberto Navigli. 2014. It’s All Fun and Games until Someone Annotates: Video Games
with a Purpose for Linguistic Annotation. Transactions of the Association for Computational Linguistics
2, 449–464.
Tobias Käfer, Ahmed Abdelrahman, Jürgen Umbrich, Patrick O’Byrne, and Aidan Hogan. 2013.
Observing Linked Data Dynamics. In The Semantic Web: Semantics and Big Data, 10th International
Conference, ESWC 2013, Montpellier, France, May 26-30, 2013. Proceedings, Philipp Cimiano, Óscar
Corcho, Valentina Presutti, Laura Hollink, and Sebastian Rudolph (Eds.). Lecture Notes in Computer
Science, vol. 7882. Springer, 213–227.
Lucie-Aimée Kaffee, Alessandro Piscopo, Pavlos Vougiouklis, Elena Simperl, Leslie Carr, and Lydia
Pintscher. 2017. A Glimpse into Babel: An Analysis of Multilinguality in Wikidata. In Proceedings of
Table of Contents
the 13th International Symposium on Open Collaboration, OpenSym 2017, Galway, Ireland, August 23-25,
2017, Lorraine Morgan (Ed.). ACM Press, 14:1–14:5.
Hans Kamp. 1981. A Theory of Truth and Semantic Representation. In Formal Methods in the Study of
Language, Jeroen A. G. Groenendijk, Theo M. V. Janssen, and Martin B. J. Stokhof (Eds.).
Mathematical Centre Tracts, vol. 136. Mathematisch Centrum Amsterdam, 277–322.
Elias Kärle, Umutcan Simsek, Oleksandra Panasiuk, and Dieter Fensel. 2018. Building an Ecosystem for
the Tyrolean Tourism Knowledge Graph. CoRR abs/1805.05744. 8 pages.
Gjergji Kasneci, Fabian M. Suchanek, Georgiana Ifrim, Maya Ramanath, and Gerhard Weikum. 2008.
NAGA: Searching and Ranking Knowledge. In Proceedings of the 24th International Conference on Data
Engineering, ICDE 2008, April 7-12, 2008, Cancún, Mexico, Gustavo Alonso, José A. Blakeley, and
Arbee L. P. Chen (Eds.). IEEE Computer Society, 953–962.
Andreas Kasten, Ansgar Scherp, Frederik Armknecht, and Matthias Krause. 2013. Towards Search on
Encrypted Graph Data. In Proceedings of the Workshop on Society, Privacy and the Semantic Web -
Policy and Technology (PrivOn2013) co-located with the 12th International Semantic Web Conference
(ISWC 2013), Sydney, Australia, October 22, 2013, Stefan Decker, Jim Hendler, and Sabrina Kirrane
(Eds.). CEUR Workshop Proceedings, vol. 1121. Sun SITE Central Europe (CEUR), 46–57.
Seyed Mehran Kazemi and David Poole. 2018. SimplE Embedding for Link Prediction in Knowledge
Graphs. In Advances in Neural Information Processing Systems 31: Annual Conference on Neural
Information Processing Systems 2018, NeurIPS 2018, 3-8 December 2018, Montréal, Canada, Samy
Bengio, Hanna M. Wallach, Hugo Larochelle, Kristen Grauman, Nicolò Cesa-Bianchi, and Roman
Garnett (Eds.), 4289–4300.
Seyed Mehran Kazemi, Rishab Goel, Kshitij Jain, Ivan Kobyzev, Akshay Sethi, Peter Forsyth, and Pascal
Poupart. 2019. Relational Representation Learning for Dynamic (Knowledge) Graphs: A Survey.
CoRR abs/1905.11485.
C. Maria Keet and Agnieszka Ławrynowicz. 2016. Test-driven development of ontologies. In The Semantic
Web. Latest Advances and New Domains - 13th International Conference, ESWC 2016, Heraklion, Crete,
Greece, May 29 - June 2, 2016, Proceedings, Harald Sack, Eva Blomqvist, Mathieu d’Aquin, Chiara
Ghidini, Simone Paolo Ponzetto, and Christoph Lange (Eds.). Lecture Notes in Computer Science,
vol. 9678. Springer, 642–657.
C. Maria Keet. 2018. An Introduction to Ontology Engineering. College Publications.
Mayank Kejriwal, Craig A. Knoblock, and Pedro Szekely (Eds.). 2021. Knowledge Graphs: Fundamentals,
Techniques, and Applications. The MIT Press.
Elisa F. Kendall and Deborah L. McGuinness. 2019. Ontology Engineering. Synthesis Lectures on the
Semantic Web: Theory and Technology, vol. 9. Morgan & Claypool.
Michael Kifer and Harold Boley. 2013. RIF Overview (Second Edition). W3C Working Group Note. World
Wide Web Consortium. February 5, 2013.
Thomas N. Kipf and Max Welling. 2017. Semi-Supervised Classification with Graph Convolutional
Networks. In 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April
24-26, 2017, Conference Track Proceedings. OpenReview.net. 14 pages.
Give us feedback!
https://kgbook.org 123/150
2/15/23, 5:13 PM Knowledge Graphs
Sabrina Kirrane, Ahmed Abdelrahman, Alessandra Mileo, and Stefan Decker. 2013. Secure Manipulation
of Linked Data. In The Semantic Web - ISWC 2013 - 12th International Semantic Web Conference,
Sydney, NSW, Australia, October 21-25, 2013, Proceedings, Part I, Harith Alani, Lalana Kagal, Achille
Fokoue, Paul Groth, Chris Biemann, Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty,
and Krzysztof Janowicz (Eds.). Lecture Notes in Computer Science, vol. 8218. Springer, 248–263.
Sabrina Kirrane, Alessandra Mileo, and Stefan Decker. 2017. Access control and the Resource Description
Framework: A survey. Semantic Web Journal 8(2), 311–352.
Jon M. Kleinberg. 1999. Hubs, authorities, and communities. ACM Computing Surveys 31(4es), p5.
Peter Kluegl, Martin Atzmueller, and Frank Puppe. 2009. TextMarker: A Tool for Rule-Based Information
Extraction. In UIMA-GSCL Workshop, 233–240.
Holger Knublauch and Dimitris Kontokostas. 2017. Shapes Constraint Language (SHACL). W3C
Recommendation. World Wide Web Consortium. June 20, 2017.
Holger Knublauch, James A. Hendler, and Kingsley Idehen. 2011. SPIN – Overview and Motivation. W3C
Member Submission. February 22, 2011.
Hanna Köpcke and Erhard Rahm. 2010. Frameworks for entity matching: A comparison. Data and
Knowledge Engineering 69(2), 197–210.
Table of Contents
Arun Krishnan. 2018. Making search easier: How Amazon’s Product Graph is helping customers find
products more easily. Amazon Blog. August 17, 2018.
Adila Krisnadhi and Pascal Hitzler. 2016. A Core Pattern for Events. In Advances in Ontology Design and
Patterns [revised and extended versions of the papers presented at the 7th edition of the Workshop on
Ontology and Semantic Web Patterns, WOP@ISWC 2016, Kobe, Japan, 18th October 2016], Karl
Hammar, Pascal Hitzler, Adila Krisnadhi, Agnieszka Lawrynowicz, Andrea Giovanni Nuzzolese, and
Monika Solanki (Eds.). Studies on the Semantic Web, vol. 32. IOS Press, 29–37.
Adila Krisnadhi and Pascal Hitzler. 2016. Modeling With Ontology Design Patterns: Chess Games As a
Worked Example. In Ontology Engineering with Ontology Design Patterns - Foundations and Applications,
Pascal Hitzler, Aldo Gangemi, Krzysztof Janowicz, Adila Krisnadhi, and Valentina Presutti (Eds.).
Studies on the Semantic Web, vol. 25. IOS Press, 3–21.
Alex Krizhevsky, Ilya Sutskever, and Geoffrey E. Hinton. 2017. ImageNet classification with deep convo‐
lutional neural networks. Communications of the ACM 60(6), 84–90.
Markus Krötzsch, Maximilian Marx, Ana Ozaki, and Veronika Thost. 2018. Attributed Description Logics:
Reasoning on Knowledge Graphs. In Proceedings of the Twenty-Seventh International Joint Conference
on Artificial Intelligence, IJCAI 2018, July 13-19, 2018, Stockholm, Sweden, Jérôme Lang (Ed.).
IJCAI/AAAI, 5309–5313.
Peter Kümmel. 1973. An Algorithm of Limited Syntax Based on Language Universals. In Computational
And Mathematical Linguistics: Proceedings of the 5th International Conference on Computational
Linguistics, COLING 1973, Pisa, Italy, August 27 - September 1, 1973, Antonio Zampolli and Nicoletta
Calzolari (Eds.). The Association for Computational Linguistics, 225–248.
H. T. Kung. 1982. Why Systolic Architectures?. IEEE Computer 15(1), 37–46.
Jose Emilio Labra Gayo, Eric Prud’hommeaux, Iovka Boneva, and Dimitris Kontokostas. 2018. Validating
RDF Data. Synthesis Lectures on the Semantic Web: Theory and Technology, vol. 7. Morgan &
Claypool.
Jose Emilio Labra Gayo, Herminio García-González, Daniel Fernández-Alvarez, and Eric Prud’hommeaux.
2019. Challenges in RDF Validation. In Current Trends in Semantic Web Technologies: Theory and
Practice, Giner Alor-Hernández, José Luis Sánchez-Cervantes, Alejandro Rodríguez-González, and
Rafael Valencia-García (Eds.). Studies in Computational Intelligence. Springer, 121–151.
Guillaume Lample, Miguel Ballesteros, Sandeep Subramanian, Kazuya Kawakami, and Chris Dyer. 2016.
Neural Architectures for Named Entity Recognition. In NAACL HLT 2016, The 2016 Conference of the
North American Chapter of the Association for Computational Linguistics: Human Language Technologies,
San Diego California, USA, June 12-17, 2016, Kevin Knight, Ani Nenkova, and Owen Rambow (Eds.).
The Association for Computational Linguistics, 260–270.
Ni Lao and William W. Cohen. 2010. Relational retrieval using a combination of path-constrained ran‐
dom walks. Machine Learning 81(1), 53–67.
Jens Lehmann, Tim Furche, Giovanni Grasso, Axel-Cyrille Ngonga Ngomo, Christian Schallhart, Andrew
Jon Sellers, Christina Unger, Lorenz Bühmann, Daniel Gerber, Konrad Höffner, David Liu, and Sören
Auer. 2012. deqa: Deep Web Extraction for Question Answering. In The Semantic Web - ISWC 2012 -
11th International Semantic Web Conference, Boston, MA, USA, November 11-15, 2012, Proceedings, Part
Give us feedback!
https://kgbook.org 124/150
2/15/23, 5:13 PM Knowledge Graphs
II, Philippe Cudré-Mauroux, Jeff Heflin, Evren Sirin, Tania Tudorache, Jérôme Euzenat, Manfred
Hauswirth, Josiane Xavier Parreira, Jim Hendler, Guus Schreiber, Abraham Bernstein, and Eva
Blomqvist (Eds.). Lecture Notes in Computer Science, vol. 7650. Springer, 131–147.
Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N. Mendes, Sebastian
Hellmann, Mohamed Morsey, Patrick van Kleef, Sören Auer, and Christian Bizer. 2015. DBpedia - A
large-scale, multilingual knowledge base extracted from Wikipedia. Semantic Web Journal 6(2), 167–
195.
Oliver Lehmberg, Dominique Ritze, Robert Meusel, and Christian Bizer. 2016. A Large Public Corpus of
Web Tables containing Time and Context Metadata. In Proceedings of the 25th International Conference
on World Wide Web, WWW 2016, Montreal, Canada, April 11-15, 2016, Companion Volume, Jacqueline
Bourdeau, Jim Hendler, Roger Nkambou, Ian Horrocks, and Ben Y. Zhao (Eds.). ACM Press, 75–76.
Yuangui Lei, Victoria Uren, and Enrico Motta. 2007. A framework for evaluating semantic metadata. In
Proceedings of the Fourth International Conference on Knowledge Capture, Derek Sleeman and Ken
Barker (Eds.). ACM Press, 135–142.
Douglas B. Lenat. 1995. CYC: A large-scale investment in knowledge infrastructure. Communications of the
ACM 38(11), 33–38.
Table of Contents
Mark Levene and Alexandra Poulovassilis. 1989. The Hypernode Model: A Graph-Theoretic Approach to
Integrating Data and Computation. In Workshop on Foundations of Models and Languages for Data and
Objects, Aigen, Austria, 25.-29. September 1989, Andreas Heuer (Ed.). Informatik-Berichte des IfI,
vol. 89-2. Technische Universität Clausthal, 55–77.
Ninghui Li, Tiancheng Li, and Suresh Venkatasubramanian. 2007. t-closeness: Privacy beyond k-anonymi‐
ty and l-diversity. In Proceedings of the 23rd International Conference on Data Engineering, ICDE 2007,
The Marmara Hotel, Istanbul, Turkey, April 15-20, 2007, Rada Chirkova, Asuman Dogac, M. Tamer
Özsu, and Timos K. Sellis (Eds.). IEEE Computer Society, 106–115.
Girija Limaye, Sunita Sarawagi, and Soumen Chakrabarti. 2010. Annotating and Searching Web Tables
Using Entities, Types and Relationships. Proceedings of the VLDB Endowment 3(1), 1338–1347.
Zhiyuan Lin and Mahesh Tripunitara. 2017. Graph Automorphism-Based, Semantics-Preserving Security
for the Resource Description Framework (RDF). In Proceedings of the Seventh ACM on Conference on
Data and Application Security and Privacy, CODASPY 2017, Scottsdale, AZ, USA, March 22-24, 2017,
Gail-Joon Ahn, Alexander Pretschner, and Gabriel Ghinita (Eds.). ACM Press, 337–348.
Yankai Lin, Zhiyuan Liu, Maosong Sun, Yang Liu, and Xuan Zhu. 2015. Learning entity and relation em‐
beddings for knowledge graph completion. In Proceedings of the Twenty-Ninth AAAI Conference on
Artificial Intelligence, January 25-30, 2015, Austin, Texas, USA, Blai Bonet and Sven Koenig (Eds.).
AAAI Press, 2181–2187.
Xiao Ling and Daniel S. Weld. 2012. Fine-Grained Entity Recognition. In Proceedings of the Twenty-Sixth
AAAI Conference on Artificial Intelligence, July 22-26, 2012, Toronto, Ontario, Canada, Jörg Hoffmann
and Bart Selman (Eds.). AAAI Press, 94–100.
Kun Liu and Evimaria Terzi. 2008. Towards identity anonymization on graphs. In Proceedings of the ACM
SIGMOD International Conference on Management of Data, SIGMOD 2008, Vancouver, BC, Canada, June
10-12, 2008, Jason Tsong-Li Wang (Ed.). ACM Press, 93–106.
Yike Liu, Tara Safavi, Abhilash Dighe, and Danai Koutra. 2018. Graph Summarization Methods and
Applications: A Survey. ACM Computing Surveys 51(3), 62:1–62:x34.
John W. Lloyd. 1984. Foundations of logic programming. Springer.
Colin Lockard, Xin Luna Dong, Prashant Shiralkar, and Arash Einolghozati. 2018. CERES: Distantly
Supervised Relation Extraction from the Semi-Structured Web. Proceedings of the VLDB Endowment
11(10), 1084–1096.
Dave Longley and Manu Sporny. 2019. RDF Dataset Normalization, A Standard RDF Dataset Normalization
Algorithm. W3C Community Group Draft Report. February 27, 2019.
Yucheng Low, Joseph Gonzalez, Aapo Kyrola, Danny Bickson, Carlos Guestrin, and Joseph M. Hellerstein.
2012. Distributed GraphLab: A Framework for Machine Learning in the Cloud. Proceedings of the
VLDB Endowment 5(8), 716–727.
Chunliang Lu, Lidong Bing, Wai Lam, Ki Chan, and Yuan Gu. 2013. Web Entity Detection for Semi-struc‐
tured Text Data Records with Unlabeled Data. International Journal of Computational Linguistics and
Applications 4(2), 135–150.
Chun Lu, Philippe Laublet, and Milan Stankovic. 2016. Travel Attractions Recommendation with
Knowledge Graphs. In Knowledge Engineering and Knowledge Management - 20th International
Give us feedback!
https://kgbook.org 125/150
2/15/23, 5:13 PM Knowledge Graphs
Conference, EKAW 2016, Bologna, Italy, November 19-23, 2016, Proceedings, Eva Blomqvist, Paolo
Ciancarini, Francesco Poggi, and Fabio Vitali (Eds.). Lecture Notes in Computer Science, vol. 10024.
Springer, 416–431.
Thomas Lukasiewicz, Maria Vanina Martinez, and Gerardo I. Simari. 2013. Complexity of Inconsistency-
Tolerant Query Answering in Datalog+/-. In Informal Proceedings of the 26th International Workshop
on Description Logics, Ulm, Germany, July 23 - 26, 2013, Thomas Eiter, Birte Glimm, Yevgeny
Kazakov, and Markus Krötzsch (Eds.). CEUR Workshop Proceedings, vol. 1014. Sun SITE Central
Europe (CEUR), 791–803.
Gang Luo, Xiaojiang Huang, Chin-Yew Lin, and Zaiqing Nie. 2015. Joint Entity Recognition and
Disambiguation. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language
Processing, EMNLP 2015, Lisbon, Portugal, September 17-21, 2015, Lluís Màrquez, Chris Callison-
Burch, Jian Su, Daniele Pighin, and Yuval Marton (Eds.). The Association for Computational
Linguistics, 879–888.
Ricardo José Machado and Armando Freitas da Rocha. 1990. The Combinatorial Neural Network: A
Connectionist Model for Knowledge Based Systems. In Uncertainty in Knowledge Bases, 3rd
International Conference on Information Processing and Management of Uncertainty in Knowledge-Based
Table of Contents
Systems, IPMU '90, Paris, France, July 2-6, 1990, Proceedings, Bernadette Bouchon-Meunier, Ronald R.
Yager, and Lotfi A. Zadeh (Eds.). Lecture Notes in Computer Science, vol. 521. Springer, 578–587.
Jayant Madhavan, David Ko, Lucja Kot, Vignesh Ganapathy, Alex Rasmussen, and Alon Y. Halevy. 2008.
Google’s Deep Web crawl. Proceedings of the VLDB Endowment 1(2), 1241–1252.
Farzaneh Mahdisoltani, Joanna Biega, and Fabian M. Suchanek. 2015. YAGO3: A Knowledge Base from
Multilingual Wikipedias. In CIDR 2015, Seventh Biennial Conference on Innovative Data Systems
Research, Asilomar, CA, USA, January 4-7, 2015, Online Proceedings. www.cidrdb.org. 11 pages.
Pierre Maillot and Carlos Bobed. 2018. Measuring structural similarity between RDF graphs. In
Proceedings of the 33rd Annual ACM Symposium on Applied Computing, SAC 2018, Pau, France, April
09-13, 2018, Hisham M. Haddad, Roger L. Wainwright, and Richard Chbeir (Eds.). ACM Press, 1960–
1967.
Grzegorz Malewicz, Matthew H. Austern, Aart J. C. Bik, James C. Dehnert, Ilan Horn, Naty Leiser, and
Grzegorz Czajkowski. 2010. Pregel: a system for large-scale graph processing. In Proceedings of the
ACM SIGMOD International Conference on Management of Data, SIGMOD 2010, Indianapolis, Indiana,
USA, June 6-10, 2010, Ahmed K. Elmagarmid and Divyakant Agrawal (Eds.). ACM Press, 135–146.
Stanislav Malyshev, Markus Krötzsch, Larry González, Julius Gonsior, and Adrian Bielefeldt. 2018.
Getting the most out of Wikidata: Semantic technology usage in Wikipedia’s knowledge graph. In The
Semantic Web - ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA, USA, October
8-12, 2018, Proceedings, Part II, Denny Vrandecic, Kalina Bontcheva, Mari Carmen Suárez-Figueroa,
Valentina Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture
Notes in Computer Science, vol. 11137. Springer, 376–394.
Ezio Marchi and Osvaldo Miguel. 1974. On the structure of the teaching-learning interactive process.
International Journal of Game Theory 3, 83–99.
Jose L. Martínez-Rodríguez, Ivan López-Arévalo, and Ana B. Rios-Alvarado. 2018. OpenIE-based ap‐
proach for Knowledge Graph construction from text. Expert Systems With Applications 113, 339–355.
Jose L. Martínez-Rodríguez, Aidan Hogan, and Ivan Lopez-Arevalo. 2020. Information Extraction meets
the Semantic Web: A Survey. Semantic Web Journal 11(2), 255–335.
Ricardo Alonso Maturana, Elena Alvarado-Cortes, Susana López-Sola, María Ortega Martínez-Losa, and
Pablo Hermoso-González. 2018. La Rioja Turismo: The Construction and Exploitation of a Queryable
Tourism Knowledge Graph. In Current Trends in Web Engineering - ICWE 2018 International Workshops,
MATWEP, EnWot, KD-WEB, WEOD, TourismKG, Cáceres, Spain, June 5, 2018, Revised Selected Papers,
Cesare Pautasso, Fernando Sánchez-Figueroa, Kari Systä, and Juan Manuel Murillo Rodriguez (Eds.).
Lecture Notes in Computer Science, vol. 11153. Springer, 213–220.
Cynthia Matuszek, John Cabral, Michael J. Witbrock, and John De Oliveira. 2006. An Introduction to the
Syntax and Content of Cyc. In Formalizing and Compiling Background Knowledge and Its Applications to
Knowledge Representation and Question Answering, Papers from the 2006 AAAI Spring Symposium,
Technical Report SS-06-05, Stanford, California, USA, March 27-29, 2006, Chitta Baral (Ed.). AAAI
Press, 44–49.
Mausam, Michael Schmitz, Stephen Soderland, Robert Bart, and Oren Etzioni. 2012. Open Language
Learning for Information Extraction. In Proceedings of the 2012 Joint Conference on Empirical Methods
Give us feedback!
https://kgbook.org 126/150
2/15/23, 5:13 PM Knowledge Graphs
in Natural Language Processing and Computational Natural Language Learning, EMNLP-CoNLL 2012, July
12-14, 2012, Jeju Island, Korea, Jun’ichi Tsujii, James Henderson, and Marius Pasça (Eds.). The
Association for Computational Linguistics, 523–534.
Mausam. 2016. Open Information Extraction Systems and Downstream Applications. In Proceedings of the
Twenty-Fifth International Joint Conference on Artificial Intelligence, IJCAI 2016, New York, NY, USA, 9-
15 July 2016, Subbarao Kambhampati (Ed.). IJCAI/AAAI, 4074–4077.
Diana Maynard, Kalina Bontcheva, and Isabelle Augenstein. 2016. Natural Language Processing for the
Semantic Web. Morgan & Claypool.
John McCarthy. 1990. Formalizing Common Sense: Papers by John McCarthy. Greenwood Publishing Group.
John McCarthy. 1993. Notes on Formalizing Context. In Proceedings of the 13th International Joint
Conference on Artificial Intelligence. Chambéry, France, August 28 - September 3, 1993, Ruzena Bajcsy
(Ed.). Morgan Kaufmann, 555–562.
Edgar Meij. 2019. Understanding News using the Bloomberg Knowledge Graph. Invited talk at the Big
Data Innovators Gathering (TheWebConf).
Pablo N. Mendes, Max Jakob, and Christian Bizer. 2012. DBpedia: A Multilingual Cross-domain
Knowledge Base. In Proceedings of the Eighth International Conference on Language Resources and
Table of Contents
Evaluation, LREC 2012, Istanbul, Turkey, May 23-25, 2012, Nicoletta Calzolari, Khalid Choukri,
Thierry Declerck, Mehmet Ugur Dogan, Bente Maegaard, Joseph Mariani, Jan Odijk, and Stelios
Piperidis (Eds.). European Language Resources Association (ELRA), 1813–1817.
Pablo N. Mendes, Hannes Mühleisen, and Christian Bizer. 2012. Sieve: linked data quality assessment and
fusion. In Proceedings of the 2012 Joint EDBT/ICDT Workshops, Berlin, Germany, March 30, 2012,
Divesh Srivastava and Ismail Ari (Eds.). ACM Press, 116–123.
Nandana Mihindukulasooriya, Mohammad Rifat Ahmmad Rashid, Giuseppe Rizzo, Raúl García-Castro,
Óscar Corcho, and Marco Torchiano. 2018. RDF shape induction using knowledge base profiling. In
Proceedings of the 33rd Annual ACM Symposium on Applied Computing, SAC 2018, Pau, France, April
09-13, 2018, Hisham M. Haddad, Roger L. Wainwright, and Richard Chbeir (Eds.). ACM Press, 1952–
1959.
Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word represen‐
tations in vector space. In 1st International Conference on Learning Representations, ICLR 2013,
Scottsdale, Arizona, USA, May 2-4, 2013, Workshop Track Proceedings, Yoshua Bengio and Yann LeCun
(Eds.). OpenReview.net. 12 pages.
George A. Miller and Christiane Fellbaum. 2007. WordNet then and now. Language Resources and
Evaluation 41(2), 209–214.
Justin J. Miller. 2013. Graph Database Applications and Concepts with Neo4j. In Proceedings of the
Southern Association for Information Systems Conference, Atlanta, GA, USA March 23rd-24th, 2013.
Association for Information Systems, 141–147.
Marvin Minsky. 1974. A Framework for representing knowledge. MIT-AI Memo 306, Santa Monica. 76
pages.
Mike Mintz, Steven Bills, Rion Snow, and Daniel Jurafsky. 2009. Distant supervision for relation extrac‐
tion without labeled data. In ACL 2009, Proceedings of the 47th Annual Meeting of the Association for
Computational Linguistics and the 4th International Joint Conference on Natural Language Processing of
the AFNLP, 2-7 August 2009, Singapore, Keh-Yih Su, Jian Su, and Janyce Wiebe (Eds.). The
Association for Computational Linguistics, 1003–1011.
Tom M. Mitchell, William W. Cohen, Estevam R. Hruschka Jr., Partha P. Talukdar, Bo Yang, Justin
Betteridge, Andrew Carlson, Bhavana Dalvi Mishra, Matt Gardner, Bryan Kisiel, Jayant
Krishnamurthy, Ni Lao, Kathryn Mazaitis, Thahir Mohamed, Ndapandula Nakashole, Emmanouil A.
Platanios, Alan Ritter, Mehdi Samadi, Burr Settles, Richard C. Wang, Derry Wijaya, Abhinav Gupta,
Xinlei Chen, Abulhair Saparov, Malcolm Greaves, and Joel Welling. 2018. Never-ending learning.
Communications of the ACM 61(5), 103–115.
Federico Monti, Davide Boscaini, Jonathan Masci, Emanuele Rodolà, Jan Svoboda, and Michael M.
Bronstein. 2017. Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs. In
2017 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, Honolulu, HI, USA, July
21-26, 2017. IEEE Computer Society, 5425–5434.
Elena Montiel-Ponsoda, Víctor Rodríguez-Doncel, and Jorge Gracia. 2017. Building the Legal Knowledge
Graph for Smart Compliance Services in Multilingual Europe. In Proceedings of the 1st Workshop on
Technologies for Regulatory Compliance co-located with the 30th International Conference on Legal
Give us feedback!
https://kgbook.org 127/150
2/15/23, 5:13 PM Knowledge Graphs
Knowledge and Information Systems (JURIX 2017), Luxembourg, December 13, 2017, Víctor Rodríguez-
Doncel, Pompeu Casanovas, and Jorge González-Conejero (Eds.). CEUR Workshop Proceedings,
vol. 2049. Sun SITE Central Europe (CEUR), 15–17.
Benjamin Moreau, Patricia Serrano-Alvarado, Matthieu Perrin, and Emmanuel Desmontils. 2019.
Modelling the Compatibility of Licenses. In The Semantic Web - 16th International Conference, ESWC
2019, Portorož, Slovenia, June 2-6, 2019, Proceedings, Pascal Hitzler, Miriam Fernández, Krzysztof
Janowicz, Amrapali Zaveri, Alasdair J. G. Gray, Vanessa López, Armin Haller, and Karl Hammar
(Eds.). Lecture Notes in Computer Science, vol. 11503. Springer, 255–269.
José Moreno-Vega and Aidan Hogan. 2018. GraFa: Scalable Faceted Browsing for RDF Graphs. In The
Semantic Web - ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA, USA, October
8-12, 2018, Proceedings, Part I, Denny Vrandečić, Kalina Bontcheva, Mari Carmen Suárez-Figueroa,
Valentina Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture
Notes in Computer Science, vol. 11136. Springer, 301–317.
Andrea Moro and Roberto Navigli. 2013. Integrating Syntactic and Semantic Analysis into the Open
Information Extraction Paradigm. In IJCAI 2013, Proceedings of the 23rd International Joint Conference
on Artificial Intelligence, Beijing, China, August 3-9, 2013, Francesca Rossi (Ed.). IJCAI/AAAI, 2148–
Table of Contents
2154.
Andrea Moro, Alessandro Raganato, and Roberto Navigli. 2014. Entity Linking meets Word Sense
Disambiguation: a Unified Approach. Transactions of the Association for Computational Linguistics 2,
231–244.
Boris Motik, Rob Shearer, and Ian Horrocks. 2009. Hypertableau Reasoning for Description Logics.
Journal of Artificial Intelligence Research 36, 165–228.
Boris Motik, Bernardo Cuenca Grau, Ian Horrocks, Zhe Wu, Achille Fokoue, and Carsten Lutz. 2012. OWL
2 Web Ontology Language Profiles (Second Edition). W3C Recommendation. World Wide Web
Consortium. December 11, 2012.
Varish Mulwad, Tim Finin, and Anupam Joshi. 2013. Semantic Message Passing for Generating Linked
Data from Tables. In The Semantic Web - ISWC 2013 - 12th International Semantic Web Conference,
Sydney, NSW, Australia, October 21-25, 2013, Proceedings, Part I, Harith Alani, Lalana Kagal, Achille
Fokoue, Paul Groth, Chris Biemann, Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty,
and Krzysztof Janowicz (Eds.). Lecture Notes in Computer Science, vol. 8218. Springer, 363–378.
Chris Mungall, Alan Ruttenberg, Ian Horrocks, and David Osumi-Sutherland. 2012. OBO Flat File Format
1.4 Syntax and Semantics. Editor's Draft. May, 2012.
Sergio Muñoz, Jorge Pérez, and Claudio Gutiérrez. 2009. Simple and Efficient Minimal RDFS. Journal of
Web Semantics 7(3), 220–234.
Emir Muñoz, Aidan Hogan, and Alessandra Mileo. 2014. Using Linked Data to mine RDF from
Wikipedia’s tables. In Seventh ACM International Conference on Web Search and Data Mining, WSDM
2014, New York, NY, USA, February 24-28, 2014, Ben Carterette, Fernando Diaz, Carlos Castillo, and
Donald Metzler (Eds.). ACM Press, 533–542.
David Nadeau and Satoshi Sekine. 2007. A survey of named entity recognition and classification.
Lingvisticæ Investigationes 30(1), 3–26.
Ndapandula Nakashole, Tomasz Tylenda, and Gerhard Weikum. 2013. Fine-grained Semantic Typing of
Emerging Entities. In Proceedings of the 51st Annual Meeting of the Association for Computational
Linguistics, ACL 2013, 4-9 August 2013, Sofia, Bulgaria, Volume 1: Long Papers. The Association for
Computational Linguistics, 1488–1497.
Arvind Narayanan and Vitaly Shmatikov. 2009. De-anonymizing Social Networks. In 30th IEEE
Symposium on Security and Privacy (S&P 2009), 17-20 May 2009, Oakland, California, USA. IEEE
Computer Society, 173–187.
Roberto Navigli and Simone Paolo Ponzetto. 2012. BabelNet: The automatic construction, evaluation and
application of a wide-coverage multilingual semantic network. Artificial Intelligence 193, 217–250.
Roberto Navigli. 2009. Word Sense Disambiguation: A Survey. ACM Computing Surveys 41(2), 1–69.
Markus Nentwig, Michael Hartung, Axel-Cyrille Ngonga Ngomo, and Erhard Rahm. 2017. A survey of
current link discovery frameworks. Semantic Web Journal 8(3), 419–436.
Sebastian Neumaier and Axel Polleres. 2019. Enabling Spatio-Temporal Search in Open Data. Journal of
Web Semantics 55, 21–36.
Sebastian Neumaier, Jürgen Umbrich, Josiane Xavier Parreira, and Axel Polleres. 2016. Multi-level
Semantic Labelling of Numerical Values. In The Semantic Web - ISWC 2016 - 15th International
Give us feedback!
https://kgbook.org 128/150
2/15/23, 5:13 PM Knowledge Graphs
Semantic Web Conference, Kobe, Japan, October 17-21, 2016, Proceedings, Part I, Paul T. Groth, Elena
Simperl, Alasdair J. G. Gray, Marta Sabou, Markus Krötzsch, Freddy Lécué, Fabian Flöck, and
Yolanda Gil (Eds.). Lecture Notes in Computer Science, vol. 9981. Springer, 428–445.
David Newman. 2019. Knowledge Graphs and AI: The Future of Financial Data.
Axel-Cyrille Ngonga Ngomo and Sören Auer. 2011. LIMES - A Time-Efficient Approach for Large-Scale
Link Discovery on the Web of Data. In IJCAI 2011, Proceedings of the 22nd International Joint
Conference on Artificial Intelligence, Barcelona, Catalonia, Spain, July 16-22, 2011, Toby Walsh (Ed.).
IJCAI/AAAI, 2312–2317.
Axel-Cyrille Ngonga Ngomo, Lorenz Bühmann, Christina Unger, Jens Lehmann, and Daniel Gerber. 2013.
Sorry, I don’t speak SPARQL: translating SPARQL queries into natural language. In 22nd International
World Wide Web Conference, WWW '13, Rio de Janeiro, Brazil, May 13-17, 2013, Daniel Schwabe,
Virgílio A. F. Almeida, Hartmut Glaser, Ricardo Baeza-Yates, and Sue B. Moon (Eds.). ACM Press,
977–988.
Axel-Cyrille Ngonga Ngomo. 2012. Link discovery with guaranteed reduction ratio in affine spaces with
Minkowski measures. In The Semantic Web - ISWC 2012 - 11th International Semantic Web Conference,
Boston, MA, USA, November 11-15, 2012, Proceedings, Part I, Philippe Cudré-Mauroux, Jeff Heflin,
Table of Contents
Evren Sirin, Tania Tudorache, Jérôme Euzenat, Manfred Hauswirth, Josiane Xavier Parreira, Jim
Hendler, Guus Schreiber, Abraham Bernstein, and Eva Blomqvist (Eds.). Lecture Notes in Computer
Science, vol. 7649. Springer, 378–393.
Axel-Cyrille Ngonga Ngomo. 2013. ORCHID–reduction-ratio-optimal computation of geo-spatial distances
for link discovery. In The Semantic Web - ISWC 2013 - 12th International Semantic Web Conference,
Sydney, NSW, Australia, October 21-25, 2013, Proceedings, Part I, Harith Alani, Lalana Kagal, Achille
Fokoue, Paul Groth, Chris Biemann, Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty,
and Krzysztof Janowicz (Eds.). Lecture Notes in Computer Science, vol. 8218. Springer, 395–410.
Vinh Nguyen, Olivier Bodenreider, and Amit Sheth. 2014. Don’t Like RDF Reification?: Making
Statements About Statements Using Singleton Property. In 23rd International World Wide Web
Conference, WWW '14, Seoul, Republic of Korea, April 7-11, 2014, Chin-Wan Chung, Andrei Z. Broder,
Kyuseok Shim, and Torsten Suel (Eds.). ACM Press, 759–770.
Dat Ba Nguyen, Martin Theobald, and Gerhard Weikum. 2016. J-NERD: Joint Named Entity Recognition
and Disambiguation with Rich Linguistic Features. Transactions of the Association for Computational
Linguistics 4, 215–229.
Maximilian Nickel and Volker Tresp. 2013. Tensor factorization for multi-relational learning. In Machine
Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2013, Prague, Czech
Republic, September 23-27, 2013, Proceedings, Part III, Hendrik Blockeel, Kristian Kersting, Siegfried
Nijssen, and Filip Zelezný (Eds.). Lecture Notes in Computer Science, vol. 8190. Springer, 617–621.
Maximilian Nickel, Kevin Murphy, Volker Tresp, and Evgeniy Gabrilovich. 2016. A Review of Relational
Machine Learning for Knowledge Graphs. Proceedings of the IEEE 104(1), 11–33.
Maximilian Nickel, Lorenzo Rosasco, and Tomaso A. Poggio. 2016. Holographic Embeddings of
Knowledge Graphs. In Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence, February
12-17, 2016, Phoenix, Arizona, USA, Dale Schuurmans and Michael P. Wellman (Eds.). AAAI Press,
1955–1961.
Natalya F. Noy and Deborah L. McGuinness. 2001. Ontology Development 101: A Guide to Creating Your
First Ontology. Stanford Knowledge Systems Laboratory. March, 2001.
Natasha Noy, Yuqing Gao, Anshu Jain, Anant Narayanan, Alan Patterson, and Jamie Taylor. 2019.
Industry-scale Knowledge Graphs: Lessons and Challenges. ACM Queue 17(2). 20 pages.
Sri Nurdiati and Cornelis Hoede. 2012. 25 Years of Development of Knowledge Graph Theory: the Results
and the Challenge. Memorandum 1876, University of Twente. September, 2012.
Ekpe Okorafor and Atish Ray. 2019. The path from data to knowledge. Accenture Applied Intelligence
Blog. June 19, 2019.
Lawrence Page, Sergey Brin, Rajeev Motwani, and Terry Winograd. 1999. The PageRank Citation Ranking:
Bringing order to the Web. Stanford InfoLab. November, 1999.
Jeff Z. Pan, Guido Vetere, José Manuél Gómez-Pérez, and Honghan Wu (Eds.). 2017. Exploiting Linked
Data and Knowledge Graphs in Large Organisations. Springer.
Oleksandra Panasiuk, Simon Steyskal, Giray Havur, Anna Fensel, and Sabrina Kirrane. 2018. Modeling
and Reasoning over Data Licenses. In The Semantic Web: ESWC 2018 Satellite Events - ESWC 2018
Satellite Events, Heraklion, Crete, Greece, June 3-7, 2018, Revised Selected Papers, Aldo Gangemi, Anna
Give us feedback!
https://kgbook.org 129/150
2/15/23, 5:13 PM Knowledge Graphs
Lisa Gentile, Andrea Giovanni Nuzzolese, Sebastian Rudolph, Maria Maleshkova, Heiko Paulheim,
Jeff Z. Pan, and Mehwish Alam (Eds.). Lecture Notes in Computer Science, vol. 11155. Springer,
218–222.
Harshvardhan J. Pandit, Axel Polleres, Beatriz Esteves, Bert Bos, Bud Bruegger, Elmar Kiesling, Eva
Schlehahn, David Hickey, Fajar J. Ekaputra, Georg P. Krog, Javier D. Fernández, Mark Lizar, Paul
Ryan, Piero Bonatti, Ramisa Gachpaz Hamed, Rigo Wenning, Rob Brennan, and Simon Steyskal.
2021. Data Privacy Vocabulary (DPV) version 0.3. Draft Community Group Report. World Wide Web
Consortium. October 1, 2021.
Vicky Papavasileiou, Giorgos Flouris, Irini Fundulaki, Dimitris Kotzinos, and Vassilis Christophides. 2013.
High-level change detection in RDF(S) KBs. ACM Transactions on Database Systems 38(1), 1:1–1:42.
Namyong Park, Andrey Kan, Xin Luna Dong, Tong Zhao, and Christos Faloutsos. 2019. Estimating Node
Importance in Knowledge Graphs Using Graph neural Networks. In Proceedings of the 25th ACM
SIGKDD International Conference on Knowledge Discovery & Data Mining, KDD 2019, Anchorage, AK,
USA, August 4-8, 2019, Ankur Teredesai, Vipin Kumar, Ying Li, Rómer Rosales, Evimaria Terzi, and
George Karypis (Eds.). ACM Press, 596–606.
Namyong Park, Andrey Kan, Xin Luna Dong, Tong Zhao, and Christos Faloutsos. 2020. MultiImport:
Table of Contents
Inferring Node Importance in a Knowledge Graph from Multiple Input Signals. In KDD '20: The 26th
ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, CA, USA, August 23-
27, 2020, Rajesh Gupta, Yan Liu, Jiliang Tang, and B. Aditya Prakash (Eds.). ACM Press, 503–512.
Jeff Pasternack and Dan Roth. 2010. Knowing What to Believe (when You Already Know Something). In
COLING 2010, 23rd International Conference on Computational Linguistics, Proceedings of the Conference,
23-27 August 2010, Beijing, China, Chu-Ren Huang and Dan Jurafsky (Eds.). Tsinghua University
Press, 877–885.
Jeff Pasternack and Dan Roth. 2011. Making Better Informed Trust Decisions with Generalized Fact-
Finding. In IJCAI 2011, Proceedings of the 22nd International Joint Conference on Artificial Intelligence,
Barcelona, Catalonia, Spain, July 16-22, 2011, Toby Walsh (Ed.). IJCAI/AAAI, 2324–2329.
Heiko Paulheim and Christian Bizer. 2013. Type inference on noisy RDF data. In The Semantic Web -
ISWC 2013 - 12th International Semantic Web Conference, Sydney, NSW, Australia, October 21-25,
2013, Proceedings, Part II, Harith Alani, Lalana Kagal, Achille Fokoue, Paul Groth, Chris Biemann,
Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty, and Krzysztof Janowicz (Eds.).
Lecture Notes in Computer Science, vol. 8219. Springer, 510–525.
Heiko Paulheim. 2017. Knowledge graph refinement: A survey of approaches and evaluation methods.
Semantic Web Journal 8(3), 489–508.
Heiko Paulheim. 2018. How much is a Triple? Estimating the Cost of Knowledge Graph Creation. In
Proceedings of the ISWC 2018 Posters & Demonstrations, Industry and Blue Sky Ideas Tracks co-located
with 17th International Semantic Web Conference (ISWC 2018), Monterey, USA, October 8th - to - 12th,
2018, Marieke van Erp, Medha Atre, Vanessa López, Kavitha Srinivas, and Carolina Fortuna (Eds.).
CEUR Workshop Proceedings, vol. 2180. Sun SITE Central Europe (CEUR). 4 pages.
Chaveevan Pechsiri and Rapepun Piriyakul. 2010. Explanation Knowledge Graph Construction Through
Causality Extraction from Texts. Journal of Computer Science and Technology 25(5), 1055–1070.
Charles S. Peirce. 1878. How to Make Our Ideas Clear. Popular Science Monthly 12, 286–302.
Tassilo Pellegrini, Giray Havur, Simon Steyskal, Oleksandra Panasiuk, Anna Fensel, Victor Mireles,
Thomas Thurner, Axel Polleres, Sabrina Kirrane, and Andrea Schönhofer. 2019. DALICC: A License
Management Framework for Digital Assets. In Proceedings of the Internationales Rechtsinformatik
Symposion (IRIS). 10 pages.
Thomas Pellissier Tanon, Denny Vrandečić, Sebastian Schaffert, Thomas Steiner, and Lydia Pintscher.
2016. From Freebase to Wikidata: The Great Migration. In Proceedings of the 25th International
Conference on World Wide Web, WWW 2016, Montreal, Canada, April 11-15, 2016, Jacqueline
Bourdeau, Jim Hendler, Roger Nkambou, Ian Horrocks, and Ben Y. Zhao (Eds.). ACM Press, 1419–
1428.
Thomas Pellissier Tanon, Daria Stepanova, Simon Razniewski, Paramita Mirza, and Gerhard Weikum.
2017. Completeness-Aware Rule Learning from Knowledge Graphs. In The Semantic Web - ISWC 2017
- 16th International Semantic Web Conference, Vienna, Austria, October 21-25, 2017, Proceedings, Part I,
Claudia d’Amato, Miriam Fernández, Valentina A. M. Tamma, Freddy Lécué, Philippe Cudré-
Mauroux, Juan F. Sequeda, Christoph Lange, and Jeff Heflin (Eds.). Lecture Notes in Computer
Science, vol. 10587. Springer, 507–525. Give us feedback!
https://kgbook.org 130/150
2/15/23, 5:13 PM Knowledge Graphs
Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. Glove: Global vectors for word rep‐
resentation. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing,
EMNLP 2014, October 25-29, 2014, Doha, Qatar, A meeting of SIGDAT, a Special Interest Group of the
ACL, Alessandro Moschitti, Bo Pang, and Walter Daelemans (Eds.). The Association for
Computational Linguistics, 1532–1543.
Silvio Peroni, David M. Shotton, and Fabio Vitali. 2017. One Year of the OpenCitations Corpus –
Releasing RDF-Based Scholarly Citation Data into the Public Domain. In The Semantic Web - ISWC
2017 - 16th International Semantic Web Conference, Vienna, Austria, October 21-25, 2017, Proceedings,
Part II, Claudia d’Amato, Miriam Fernández, Valentina A. M. Tamma, Freddy Lécué, Philippe Cudré-
Mauroux, Juan F. Sequeda, Christoph Lange, and Jeff Heflin (Eds.). Lecture Notes in Computer
Science, vol. 10588. Springer, 184–192.
Silvio Peroni. 2016. A simplified agile methodology for ontology development. In OWL: - Experiences and
Directions - Reasoner Evaluation - 13th International Workshop, OWLED 2016, and 5th International
Workshop, ORE 2016, Bologna, Italy, November 20, 2016, Revised Selected Papers, Mauro Dragoni,
María Poveda-Villalón, and Ernesto Jiménez-Ruiz (Eds.). Lecture Notes in Computer Science,
vol. 10161. Springer, 55–69.
Table of Contents
David Peterson, Shudi Gao, Ashok Malhotra, C. M. Sperberg-McQueen, Henry S. Thompson, and Paul V.
Biron. 2012. W3C XML Schema Definition Language (XSD) 1.1 Part 2: Datatypes. W3C
Recommendation. World Wide Web Consortium. April 5, 2012.
Giulio Petrucci, Chiara Ghidini, and Marco Rospocher. 2016. Ontology learning in the deep. In Knowledge
Engineering and Knowledge Management - 20th International Conference, EKAW 2016, Bologna, Italy,
November 19-23, 2016, Proceedings, Eva Blomqvist, Paolo Ciancarini, Francesco Poggi, and Fabio
Vitali (Eds.). Lecture Notes in Computer Science, vol. 10024. Springer, 480–495.
Minh-Duc Pham, Linnea Passing, Orri Erling, and Peter A. Boncz. 2015. Deriving an Emergent Relational
Schema from RDF Data. In Proceedings of the 24th International Conference on World Wide Web, WWW
2015, Florence, Italy, May 18-22, 2015, Aldo Gangemi, Stefano Leonardi, and Alessandro Panconesi
(Eds.). ACM Press, 864–874.
H. Sofia Pinto, C. Tempich, and Steffen Staab. 2009. Ontology Engineering and Evolution in a Distributed
World Using DILIGENT. In Handbook on Ontologies, Steffen Staab and Rudi Studer (Eds.).
International Handbooks on Information Systems. Springer, 153–176.
Alessandro Piscopo and Elena Simperl. 2018. Who Models the World?: Collaborative Ontology Creation
and User Roles in Wikidata. Proceedings of the ACM on Human-Computer Interaction 2(CSCW), 141:1–
141:18.
Alessandro Piscopo, Lucie-Aimée Kaffee, Chris Phethean, and Elena Simperl. 2017. Provenance
Information in a Collaborative Knowledge Graph: An Evaluation of Wikidata External References. In
The Semantic Web - ISWC 2017 - 16th International Semantic Web Conference, Vienna, Austria, October
21-25, 2017, Proceedings, Part I, Claudia d’Amato, Miriam Fernández, Valentina A. M. Tamma,
Freddy Lécué, Philippe Cudré-Mauroux, Juan F. Sequeda, Christoph Lange, and Jeff Heflin (Eds.).
Lecture Notes in Computer Science, vol. 10587. Springer, 542–558.
R. J. Pittman, Amit Srivastava, Sanjika Hewavitharana, Ajinkya Kale, and Saab Mansour. 2017. Cracking
the Code on Conversational Commerce. eBay Blog. April 6, 2017.
Aleksander Pivk, Philipp Cimiano, York Sure, Matjaz Gams, Vladislav Rajkovic, and Rudi Studer. 2007.
Transforming arbitrary tables into logical form with TARTAR. Data and Knowledge Engineering 60(3),
567–595.
Roel Popping. 2003. Knowledge Graphs and Network Text Analysis. Social Science Information 42(91),
91–106.
Valentina Presutti, Enrico Daga, Aldo Gangemi, and Eva Blomqvist. 2009. eXtreme Design with Content
Ontology Design Patterns. In Proceedings of the Workshop on Ontology Patterns (WOP 2009) , collocat‐
ed with the 8th International Semantic Web Conference ( ISWC-2009 ), Washington D.C., USA, 25
October, 2009, Eva Blomqvist, Kurt Sandkuhl, François Scharffe, and Vojtech Svátek (Eds.). CEUR
Workshop Proceedings, vol. 516. Sun SITE Central Europe (CEUR). 15 pages.
Eric Prud’hommeaux and Gavin Carothers. 2014. RDF 1.1 Turtle – Terse RDF Triple Language, W3C
Recommendation 25 February 2014. W3C Recommendation. World Wide Web Consortium.
February 25, 2014.
Eric Prud’hommeaux, Jose Emilio Labra Gayo, and Harold Solbrig. 2014. Shape Expressions: An RDF
Validation and Transformation Language. In Proceedings of the 10th International Conference on
Give us feedback!
https://kgbook.org 131/150
2/15/23, 5:13 PM Knowledge Graphs
Semantic Systems, SEMANTICS 2014, Leipzig, Germany, September 4-5, 2014, Harald Sack, Agata
Filipowska, Jens Lehmann, and Sebastian Hellmann (Eds.). ACM Press, 32–40.
Jay Pujara, Hui Miao, Lise Getoor, and William W. Cohen. 2013. Knowledge Graph Identification. In The
Semantic Web - ISWC 2013 - 12th International Semantic Web Conference, Sydney, NSW, Australia,
October 21-25, 2013, Proceedings, Part I, Harith Alani, Lalana Kagal, Achille Fokoue, Paul Groth,
Chris Biemann, Josian Xavier Parreira, Lora Aroyo, Natasha Noy, Chris Welty, and Krzysztof
Janowicz (Eds.). Lecture Notes in Computer Science, vol. 8218. Springer, 542–557.
Guilin Qi, Huajun Chen, Kang Liu, Haofen Wang, Qiu Ji, and Tianxing Wu. 2021. Knowledge Graph.
Database Management & Information Retrieval. Springer.
Ross Quillian. 1963. A notation for representing conceptual information: An application to semantics and me‐
chanical English paraphrasing. Systems Development Corp..
Stephan Rabanser, Oleksandr Shchur, and Stephan Günnemann. 2017. Introduction to Tensor
Decompositions and their Applications in Machine Learning. CoRR abs/1711.10781. 13 pages.
Roy Rada. 1986. Gradualness eases refinement of medical knowledge. Medical Informatics 11(1), 59–73.
Filip Radulovic, Raúl García-Castro, and Asunción Gómez-Pérez. 2015. Towards the Anonymisation of
RDF Data. In The 27th International Conference on Software Engineering and Knowledge Engineering,
Table of Contents
SEKE 2015, Wyndham Pittsburgh University Center, Pittsburgh, PA, USA, July 6-8, 2015, Haiping Xu
(Ed.). KSI Research Inc. and Knowledge Systems Institute Graduate School, 646–651.
Yves Raimond, Christopher Sutton, and Mark B. Sandler. 2009. Interlinking Music-Related Data on the
Web. IEEE MultiMedia 16(2), 52–63.
Yves Raimond, Tristan Ferne, Michael Smethurst, and Gareth Adams. 2014. The BBC World Service
Archive prototype. Journal of Web Semantics 27–28, 2–9.
Alain T. Rappaport and Albert M. Gouyet. 1988. Dynamic, interactive display system for a knowledge
base. US Patent US4752889A. June 21, 1988.
Lev-Arie Ratinov and Dan Roth. 2009. Design Challenges and Misconceptions in Named Entity
Recognition. In Proceedings of the Thirteenth Conference on Computational Natural Language Learning,
CoNLL 2009, Boulder, Colorado, USA, June 4-5, 2009, Suzanne Stevenson and Xavier Carreras (Eds.).
The Association for Computational Linguistics, 147–155.
Pavan Reddivari, Tim Finin, and Anupam Joshi. 2005. Policy-based access control for an RDF store. In
Policy Management for the Web, A workshop held at the 14th International World Wide Web Conference
Tuesday 10 May 2005, Chiba Japan, Lalana Kagal, Tim Finin, and James Hendler (Eds.), 78–81.
Raymond Reiter. 1987. A Theory of Diagnosis from First Principles. Artificial Intelligence 32(1), 57–95.
Xiang Ren, Ahmed El-Kishky, Chi Wang, Fangbo Tao, Clare R. Voss, and Jiawei Han. 2015. ClusType:
Effective Entity Recognition and Typing by Relation Phrase-Based Clustering. In Proceedings of the
21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Sydney, NSW,
Australia, August 10-13, 2015, Longbing Cao, Chengqi Zhang, Thorsten Joachims, Geoffrey I. Webb,
Dragos D. Margineantu, and Graham Williams (Eds.). ACM Press, 995–1004.
Xiang Ren, Zeqiu Wu, Wenqi He, Meng Qu, Clare R. Voss, Heng Ji, Tarek F. Abdelzaher, and Jiawei Han.
2017. CoType: Joint Extraction of Typed Entities and Relations with Knowledge Bases. In Proceedings
of the 26th International Conference on World Wide Web, WWW 2017, Perth, Australia, April 3-7, 2017,
Rick Barrett, Rick Cummings, Eugene Agichtein, and Evgeniy Gabrilovich (Eds.). ACM Press, 1015–
1024.
Juan L. Reutter, Adrián Soto, and Domagoj Vrgoc. 2015. Recursion in SPARQL. In The Semantic Web -
ISWC 2015 - 14th International Semantic Web Conference, Bethlehem, PA, USA, October 11-15, 2015,
Proceedings, Part I, Marcelo Arenas, Óscar Corcho, Elena Paslaru Bontas Simperl, Markus Strohmaier,
Mathieu d’Aquin, Kavitha Srinivas, Paul T. Groth, Michel Dumontier, Jeff Heflin, Krishnaprasad
Thirunarayan, and Stephen Staab (Eds.). Lecture Notes in Computer Science, vol. 9366. Springer,
19–35.
Sebastian Riedel, Limin Yao, and Andrew McCallum. 2010. Modeling Relations and Their Mentions with‐
out Labeled Text. In Machine Learning and Knowledge Discovery in Databases, European Conference,
ECML PKDD 2010, Barcelona, Spain, September 20-24, 2010, Proceedings, Part III, José L. Balcázar,
Francesco Bonchi, Aristides Gionis, and Michèle Sebag (Eds.). Lecture Notes in Computer Science,
vol. 6323. Springer, 148–163.
Petar Ristoski and Heiko Paulheim. 2016. RDF2Vec: RDF Graph Embeddings for Data Mining. In The
Semantic Web - ISWC 2016 - 15th International Semantic Web Conference, Kobe, Japan, October 17-21,
2016, Proceedings, Part I, Paul T. Groth, Elena Simperl, Alasdair J. G. Gray, Marta Sabou, Markus
Give us feedback!
https://kgbook.org 132/150
2/15/23, 5:13 PM Knowledge Graphs
Krötzsch, Freddy Lécué, Fabian Flöck, and Yolanda Gil (Eds.). Lecture Notes in Computer Science,
vol. 9981. Springer, 498–514.
Richard H. Richens. 1958. Interlingual Machine Translation. The Computer Journal 1(3), 144–147.
Giuseppe Rizzo, Claudia d’Amato, Nicola Fanizzi, and Floriana Esposito. 2017. Terminological Cluster
Trees for Disjointness Axiom Discovery. In The Semantic Web - 14th International Conference, ESWC
2017, Portorož, Slovenia, May 28 - June 1, 2017, Proceedings, Part I, Eva Blomqvist, Diana Maynard,
Aldo Gangemi, Rinke Hoekstra, Pascal Hitzler, and Olaf Hartig (Eds.). Lecture Notes in Computer
Science, vol. 10249. Springer, 184–201.
Giuseppe Rizzo, Nicola Fanizzi, and Claudia d’Amato. 2020. Class expression induction as concept space
exploration: From DL-Foil to DL-Focl. Future Generation Computer Systems 108, 256–272.
Giuseppe Rizzo, Claudia d’Amato, and Nicola Fanizzi. 2021. An unsupervised approach to disjointness
learning based on terminological cluster trees. Semantic Web Journal 12(3), 423–447.
Tim Rocktäschel and Sebastian Riedel. 2017. End-to-end Differentiable Proving. In Advances in Neural
Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017,
4–9 December 2017, Long Beach, CA, USA, Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna
M. Wallach, Rob Fergus, S. V. N. Vishwanathan, and Roman Garnett (Eds.), 3788–3800.
Table of Contents
Marko A. Rodriguez. 2015. The Gremlin graph traversal machine and language. In Proceedings of the 15th
Symposium on Database Programming Languages, Pittsburgh, PA, USA, October 25-30, 2015, James
Cheney and Thomas Neumann (Eds.). ACM Press, 1–10.
Stephen Roller, Douwe Kiela, and Maximilian Nickel. 2018. Hearst Patterns Revisited: Automatic
Hypernym Detection from Large Text Corpora. In Proceedings of the 56th Annual Meeting of the
Association for Computational Linguistics, ACL 2018, Melbourne, Australia, July 15-20, 2018, Volume 2:
Short Papers, Iryna Gurevych and Yusuke Miyao (Eds.). The Association for Computational
Linguistics, 358–363.
Marco Rospocher, Marieke van Erp, Piek Vossen, Antske Fokkens, Itziar Aldabe, German Rigau, Aitor
Soroa, Thomas Ploeger, and Tessel Bogaard. 2016. Building event-centric knowledge graphs from
news. Journal of Web Semantics 37–38, 132–151.
Jacobo Rouces, Gerard de Melo, and Katja Hose. 2015. Framebase: Representing $n$-ary relations using
semantic frames. In The Semantic Web. Latest Advances and New Domains - 12th European Semantic
Web Conference, ESWC 2015, Portoroz, Slovenia, May 31 - June 4, 2015. Proceedings, Fabien Gandon,
Marta Sabou, Harald Sack, Claudia d’Amato, Philippe Cudré-Mauroux, and Antoine Zimmermann
(Eds.). Lecture Notes in Computer Science, vol. 9088. Springer, 505–521.
Sebastian Rudolph, Markus Krötzsch, and Pascal Hitzler. 2008. Description Logic Reasoning with
Decision Diagrams: Compiling SHIQ to Disjunctive Datalog. In The Semantic Web - ISWC 2008, 7th
International Semantic Web Conference, ISWC 2008, Karlsruhe, Germany, October 26-30, 2008.
Proceedings, Amit P. Sheth, Steffen Staab, Mike Dean, Massimo Paolucci, Diana Maynard, Timothy W.
Finin, and Krishnaprasad Thirunarayan (Eds.). Lecture Notes in Computer Science, vol. 5318.
Springer, 435–450.
Anisa Rula, Matteo Palmonari, Andreas Harth, Steffen Stadtmüller, and Andrea Maurino. 2012. On the
Diversity and Availability of Temporal Information in Linked Open Data. In The Semantic Web - ISWC
2012 - 11th International Semantic Web Conference, Boston, MA, USA, November 11-15, 2012,
Proceedings, Part I, Philippe Cudré-Mauroux, Jeff Heflin, Evren Sirin, Tania Tudorache, Jérôme
Euzenat, Manfred Hauswirth, Josiane Xavier Parreira, Jim Hendler, Guus Schreiber, Abraham
Bernstein, and Eva Blomqvist (Eds.). Lecture Notes in Computer Science, vol. 7649. Springer, 492–
507.
Anisa Rula, Luca Panziera, Matteo Palmonari, and Andrea Maurino. 2014. Capturing the Currency of
DBpedia Descriptions and Get Insight into their Validity. In Proceedings of the 5th International
Workshop on Consuming Linked Data, COLD 2014 co-located with the 13th International Semantic Web
Conference (ISWC 2014), Riva del Garda, Italy, October 20, 2014, Olaf Hartig, Aidan Hogan, and Juan
F. Sequeda (Eds.). CEUR Workshop Proceedings, vol. 1264. Sun SITE Central Europe (CEUR). 12
pages.
Anisa Rula, Matteo Palmonari, Simone Rubinacci, Axel-Cyrille Ngonga Ngomo, Jens Lehmann, Andrea
Maurino, and Diego Esteves. 2019. TISCO: Temporal scoping of facts. Journal of Web Semantics 54,
72–86.
Owen Sacco and Alexandre Passant. 2011. A Privacy Preference Ontology (PPO) for Linked Data. In
WWW2011 Workshop on Linked Data on the Web, Hyderabad, India, March 29, 2011, Christian Bizer,
Give us feedback!
https://kgbook.org 133/150
2/15/23, 5:13 PM Knowledge Graphs
Tom Heath, Tim Berners-Lee, and Michael Hausenblas (Eds.). CEUR Workshop Proceedings, vol. 813.
Sun SITE Central Europe (CEUR). 5 pages.
Ali Sadeghian, Mohammadreza Armandpour, Patrick Ding, and Patrick Wang. 2019. DRUM: End-To-End
Differentiable Rule Mining On Knowledge Graphs. In Advances in Neural Information Processing
Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, 8–14
December 2019, Vancouver, BC, Canada, Hanna M. Wallach, Hugo Larochelle, Alina Beygelzimer,
Florence d’Alché-Buc, Emily B. Fox, and Roman Garnett (Eds.), 15321–15331.
Tomás Sáez and Aidan Hogan. 2018. Automatically Generating Wikipedia Info-boxes from Wikidata. In
Companion of the The Web Conference 2018 on The Web Conference 2018, WWW 2018, Lyon , France,
April 23-27, 2018, Pierre-Antoine Champin, Fabien L. Gandon, Mounia Lalmas, and Panagiotis G.
Ipeirotis (Eds.). ACM Press, 1823–1830.
Muhammad Saleem, Muhammad Intizar Ali, Aidan Hogan, Qaiser Mehmood, and Axel-Cyrille Ngonga
Ngomo. 2015. LSQ: The Linked SPARQL Queries Dataset. In The Semantic Web - ISWC 2015 - 14th
International Semantic Web Conference, Bethlehem, PA, USA, October 11-15, 2015, Proceedings, Part II,
Marcelo Arenas, Óscar Corcho, Elena Paslaru Bontas Simperl, Markus Strohmaier, Mathieu d’Aquin,
Kavitha Srinivas, Paul T. Groth, Michel Dumontier, Jeff Heflin, Krishnaprasad Thirunarayan, and
Table of Contents
Stephen Staab (Eds.). Lecture Notes in Computer Science, vol. 9367. Springer, 261–269.
Mehdi Samadi, Partha Talukdar, Manuela Veloso, and Manuel Blum. 2016. ClaimEval: Integrated and
Flexible Framework for Claim Evaluation Using Credibility of Sources. In Proceedings of the Thirtieth
AAAI Conference on Artificial Intelligence, February 12-17, 2016, Phoenix, Arizona, USA, Dale
Schuurmans and Michael P. Wellman (Eds.). AAAI Press, 222–228.
Pierangela Samarati and Latanya Sweeney. 1998. Protecting privacy when disclosing information: k-anonymi‐
ty and its enforcement through generalization and suppression. Computer Science Laboratory, SRI
International.
Georgios M. Santipantakis, Apostolos Glenis, Christos Doulkeridis, Akrivi Vlachou, and George A. Vouros.
2019. stLD: towards a spatio-temporal link discovery framework. In Proceedings of the International
Workshop on Semantic Big Data, SBD@SIGMOD 2019, Amsterdam, The Netherlands, July 5, 2019, Sven
Groppe and Le Gruenwald (Eds.). ACM Press, 4:1–4:6.
Eugene Santos Jr. and Eugene S. Santos. 1999. A framework for building knowledge-bases under uncer‐
tainty. Journal of Experimental & Theoretical Artificial Intelligence 11(2), 265–286.
Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. 2009. The
Graph Neural Network Model. IEEE Transactions on Neural Networks 20(1), 61–80.
Max Schmachtenberg, Christian Bizer, and Heiko Paulheim. 2014. Adoption of the Linked Data Best
Practices in Different Topical Domains. In The Semantic Web - ISWC 2014 - 13th International Semantic
Web Conference, Riva del Garda, Italy, October 19-23, 2014. Proceedings, Part I, Peter Mika, Tania
Tudorache, Abraham Bernstein, Christopher A. Welty, Craig A. Knoblock, Denny Vrandecic, Paul T.
Groth, Natasha Fridman Noy, Krzysztof Janowicz, and Carole A. Goble (Eds.). Lecture Notes in
Computer Science, vol. 8796. Springer, 245–260.
Manfred Schmidt-Schauß and Gert Smolka. 1991. Attributive Concept Descriptions with Complements.
Artificial Intelligence 48(1), 1–26.
Michael Schneider and Geoff Sutcliffe. 2011. Reasoning in the OWL 2 Full Ontology Language Using
First-Order Automated Theorem Proving. In Automated Deduction - CADE-23 - 23rd International
Conference on Automated Deduction, Wroclaw, Poland, July 31 - August 5, 2011. Proceedings, Nikolaj
Bjørner and Viorica Sofronie-Stokkermans (Eds.). Lecture Notes in Computer Science, vol. 6803.
Springer, 461–475.
E. W. Schneider. 1973. Course Modularization Applied: The Interface System and Its Implications For Sequence
Control and Data Analysis. Human Resources Research Organization, Alexandria, VA. November,
1973.
Christoph Schuetz, Loris Bozzato, Bernd Neumayr, Michael Schrefl, and Luciano Serafini. 2021.
Knowledge Graph OLAP: A Multidimensional Model and Query Operations for Contextualized
Knowledge Graphs. Semantic Web Journal 12(4).
Philipp Seifer, Johannes Härtel, Martin Leinberger, Ralf Lämmel, and Steffen Staab. 2019. Empirical
study on the usage of graph query languages in open source Java projects. In Proceedings of the 12th
ACM SIGPLAN International Conference on Software Language Engineering, SLE 2019, Athens, Greece,
October 20-22, 2019, Oscar Nierstrasz, Jeff Gray, and Bruno C. d. S. Oliveira (Eds.). ACM Press, 152–
166. Give us feedback!
https://kgbook.org 134/150
2/15/23, 5:13 PM Knowledge Graphs
Juan F. Sequeda, Marcelo Arenas, and Daniel P. Miranker. 2012. On directly mapping relational databas‐
es to RDF and OWL. In Proceedings of the 21st World Wide Web Conference 2012, WWW 2012, Lyon,
France, April 16-20, 2012, Alain Mille, Fabien L. Gandon, Jacques Misselis, Michael Rabinovich, and
Steffen Staab (Eds.). ACM Press, 649–658.
Juan F. Sequeda, Marcelo Arenas, and Daniel P. Miranker. 2014. OBDA: Query Rewriting or
Materialization? In Practice, Both!. In The Semantic Web - ISWC 2014 - 13th International Semantic
Web Conference, Riva del Garda, Italy, October 19-23, 2014. Proceedings, Part I, Peter Mika, Tania
Tudorache, Abraham Bernstein, Christopher A. Welty, Craig A. Knoblock, Denny Vrandecic, Paul T.
Groth, Natasha Fridman Noy, Krzysztof Janowicz, and Carole A. Goble (Eds.). Lecture Notes in
Computer Science, vol. 8796. Springer, 535–551.
Juan F. Sequeda, Willard J. Briggs, Daniel P. Miranker, and Wayne P. Heideman. 2019. A Pay-as-you-go
Methodology to Design and Build Enterprise Knowledge Graphs from Relational Databases. In The
Semantic Web - ISWC 2019 - 18th International Semantic Web Conference, Auckland, New Zealand,
October 26-30, 2019, Proceedings, Part II, Chiara Ghidini, Olaf Hartig, Maria Maleshkova, Vojtech
Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime Lefrançois, and Fabien Gandon (Eds.).
Lecture Notes in Computer Science, vol. 11779. Springer, 526–545.
Table of Contents
Luciano Serafini and Martin Homola. 2012. Contextualized Knowledge Repositories for the Semantic
Web. Journal of Web Semantics 12, 64–87.
Stephan Seufert, Patrick Ernst, Srikanta J. Bedathur, Sarath Kumar Kondreddi, Klaus Berberich, and
Gerhard Weikum. 2016. Instant Espresso: Interactive Analysis of Relationships in Knowledge Graphs.
In Proceedings of the 25th International Conference on World Wide Web, WWW 2016, Montreal, Canada,
April 11-15, 2016, Companion Volume, Jacqueline Bourdeau, Jim Hendler, Roger Nkambou, Ian
Horrocks, and Ben Y. Zhao (Eds.). ACM Press, 251–254.
Nigel Shadbolt and Kieron O’Hara. 2013. Linked Data in Government. IEEE Inteternet Computing 17(4),
72–77.
Mohamed Ahmed Sherif and Axel-Cyrille Ngonga Ngomo. 2015. Semantic Quran. Semantic Web Journal
6(4), 339–345.
Mohamed Ahmed Sherif and Axel-Cyrille Ngonga Ngomo. 2018. A systematic survey of point set distance
measures for link discovery. Semantic Web Journal 9(5), 589–604.
Baoxu Shi and Tim Weninger. 2016. Discriminative predicate path mining for fact checking in knowledge
graphs. Knowledge-based Systems 104, 123–133.
Cogan Shimizu, Quinn Hirt, and Pascal Hitzler. 2019. MODL: A Modular Ontology Design Library. CoRR
abs/1904.05405. 12 pages.
Solomon Eyal Shimony, Carmel Domshlak, and Eugene Santos Jr.. 1997. Cost-Sharing in Bayesian
Knowledge Bases. In UAI '97: Proceedings of the Thirteenth Conference on Uncertainty in Artificial
Intelligence, Brown University, Providence, Rhode Island, USA, August 1-3, 1997, Dan Geiger and
Prakash P. Shenoy (Eds.). Morgan Kaufmann, 421–428.
Prashant Shiralkar, Alessandro Flammini, Filippo Menczer, and Giovanni Luca Ciampaglia. 2017. Finding
streams in knowledge graphs to support fact checking. In 2017 IEEE International Conference on Data
Mining, ICDM 2017, New Orleans, LA, USA, November 18-21, 2017, Vijay Raghavan, Srinivas Aluru,
George Karypis, Lucio Miele, and Xindong Wu (Eds.). IEEE Computer Society, 859–864.
Saurabh Shrivastava. 2017. Bring rich knowledge of people, places, things and local businesses to your
apps. Bing Blogs. July 12, 2017.
Rôney Reis C. Silva, Bruno C. Leal, Felipe T. Brito, Vânia M. P. Vidal, and Javam C. Machado. 2017. A
Differentially Private Approach for Querying RDF Data of Social Networks. In Proceedings of the 21st
International Database Engineering & Applications Symposium, IDEAS 2017, Bristol, United Kingdom, July
12-14, 2017, Bipin C. Desai, Jun Hong, and Richard McClatchey (Eds.). ACM Press, 74–81.
Amit Singhal. 2012. Introducing the Knowledge Graph: things, not strings. Google Blog. May 16, 2012.
Martin G. Skjæveland, Daniel P. Lupp, Leif Harald Karlsen, and Henrik Forssell. 2018. Practical Ontology
Pattern Instantiation, Discovery, and Maintenance with Reasonable Ontology Templates. In The
Semantic Web - ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA, USA, October
8-12, 2018, Proceedings, Part I, Denny Vrandečić, Kalina Bontcheva, Mari Carmen Suárez-Figueroa,
Valentina Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture
Notes in Computer Science, vol. 11136. Springer, 477–494.
Jennifer Sleeman and Tim Finin. 2013. Type Prediction for Efficient Coreference Resolution in
Heterogeneous Semantic Graphs. In 2013 IEEE Seventh International Conference on Semantic
Give us feedback!
https://kgbook.org 135/150
2/15/23, 5:13 PM Knowledge Graphs
Computing, Irvine, CA, USA, September 16-18, 2013. IEEE Computer Society, 78–85.
Alisa Smirnova and Philippe Cudré-Mauroux. 2019. Relation Extraction Using Distant Supervision: A
Survey. ACM Computing Surveys 51(5), 106:1–106:35.
Richard Socher, Danqi Chen, Christopher D Manning, and Andrew Ng. 2013. Reasoning with neural ten‐
sor networks for knowledge base completion. In Advances in Neural Information Processing Systems 26:
27th Annual Conference on Neural Information Processing Systems 2013. Proceedings of a meeting held
December 5-8, 2013, Lake Tahoe, Nevada, United States, Christopher J. C. Burges, Léon Bottou, Zoubin
Ghahramani, and Kilian Q. Weinberger (Eds.), 926–934.
Arnaud Soulet, Arnaud Giacometti, Béatrice Markhoff, and Fabian M. Suchanek. 2018.
Representativeness of Knowledge Bases with the Generalized Benford’s Law. In The Semantic Web -
ISWC 2018 - 17th International Semantic Web Conference, Monterey, CA, USA, October 8-12, 2018,
Proceedings, Part I, Denny Vrandečić, Kalina Bontcheva, Mari Carmen Suárez-Figueroa, Valentina
Presutti, Irene Celino, Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture Notes in
Computer Science, vol. 11136. Springer, 374–390.
John Sowa. 1979. Semantics of Conceptual Graphs. In 17th Annual Meeting of the Association for
Computational Linguistics, 29 June - 1 July 1979, University of California at San Diego, La Jolla, CA,
Table of Contents
USA, Norman K. Sondheimer (Ed.). The Association for Computational Linguistics, 39–44.
John Sowa. 1987. Semantic Networks. In Encyclopedia of Cognitive Science, Stuart C. Shapiro (Ed.). John
Wiley & Sons.
Blerina Spahiu, Riccardo Porrini, Matteo Palmonari, Anisa Rula, and Andrea Maurino. 2016. ABSTAT:
Ontology-Driven Linked Data Summaries with Pattern Minimalization. In The Semantic Web - ESWC
2016 Satellite Events, Heraklion, Crete, Greece, May 29 - June 2, 2016, Revised Selected Papers, Harald
Sack, Giuseppe Rizzo, Nadine Steinmetz, Dunja Mladenic, Sören Auer, and Christoph Lange (Eds.).
Lecture Notes in Computer Science, vol. 9989. Springer, 381–395.
Alessandro Sperduti and Antonina Starita. 1997. Supervised neural networks for the classification of
structures. IEEE Transactions on Neural Networks 8(3), 714–735.
Manu Sporny, Gregg Kellogg, Markus Lanthaler, Dave Longley, and Niklas Lindström. 2014. JSON-LD 1.0,
A JSON-based Serialization for Linked Data, W3C Recommendation 16 January 2014. W3C
Recommendation. World Wide Web Consortium. January 16, 2014.
Rajan Srikanth and Matthias Jarke. 1989. The Design of Knowledge-Based Systems for Managing Ill-
Structured Software Projects. Decision Support Systems 5(4), 425–447.
Claus Stadler, Jens Lehmann, Konrad Höffner, and Sören Auer. 2012. LinkedGeoData: A core for a web of
spatial open data. Semantic Web Journal 3(4), 333–354.
Simon Steyskal and Axel Polleres. 2014. Defining expressive access policies for linked data using the
ODRL ontology 2.0. In Proceedings of the 10th International Conference on Semantic Systems,
SEMANTICS 2014, Leipzig, Germany, September 4-5, 2014, Harald Sack, Agata Filipowska, Jens
Lehmann, and Sebastian Hellmann (Eds.). ACM Press, 20–23.
Patrick Stickler. 2005. CBD – Concise Bounded Description, W3C Member Submission 3 June 2005. W3C
Member Submission. June 3, 2005.
Radu Stoica, George H. L. Fletcher, and Juan F. Sequeda. 2019. On Directly Mapping Relational
Databases to Property Graphs. In Proceedings of the 13th Alberto Mendelzon International Workshop on
Foundations of Data Management, Asunción, Paraguay, June 3-7, 2019, Aidan Hogan and Tova Milo
(Eds.). CEUR Workshop Proceedings, vol. 2369. Sun SITE Central Europe (CEUR). 4 pages.
Frans N. Stokman and Pieter H. de Vries. 1988. Structuring knowledge in a graph. In Human-Computer
Interaction, Gerrit C. van der Veer and Gijsbertus Mulder (Eds.). Springer, 186–206.
Umberto Straccia. 2009. A Minimal Deductive System for General Fuzzy RDF. In Web Reasoning and Rule
Systems, Third International Conference, RR 2009, Chantilly, VA, USA, October 25-26, 2009,
Proceedings, Axel Polleres and Terrance Swift (Eds.). Lecture Notes in Computer Science, vol. 5837.
Springer, 166–181.
Philip Stutz, Daniel Strebel, and Abraham Bernstein. 2016. Signal/Collect12. Semantic Web Journal 7(2),
139–166.
Fabian M. Suchanek, Gjergji Kasneci, and Gerhard Weikum. 2007. YAGO: A core of semantic knowledge.
In Proceedings of the 16th International Conference on World Wide Web, WWW 2007, Banff, Alberta,
Canada, May 8-12, 2007, Carey L. Williamson, Mary Ellen Zurko, Peter F. Patel-Schneider, and
Prashant J. Shenoy (Eds.). ACM Press, 697–706.
Give us feedback!
https://kgbook.org 136/150
2/15/23, 5:13 PM Knowledge Graphs
Fabian M. Suchanek, Gjergji Kasneci, and Gerhard Weikum. 2008. YAGO: A Large Ontology from
Wikipedia and WordNet. Journal of Web Semantics 6(3), 203–217.
Fabian M. Suchanek, Jonathan Lajus, Armand Boschin, and Gerhard Weikum. 2019. Knowledge
Representation and Rule Mining in Entity-Centric Knowledge Bases. In Reasoning Web. Explainable
Artificial Intelligence - 15th International Summer School 2019, Bolzano, Italy, September 20-24, 2019,
Tutorial Lectures, Markus Krötzsch and Daria Stepanova (Eds.). Lecture Notes in Computer Science,
vol. 11810. Springer, 110–152.
Yizhou Sun and Jiawei Han. 2012. Mining Heterogeneous Information Networks: Principles and
Methodologies. Synthesis Lectures on Data Mining and Knowledge Discovery, vol. 3. Morgan &
Claypool.
Yizhou Sun, Jiawei Han, Xifeng Yan, Philip S. Yu, and Tianyi Wu. 2011. Pathsim: Meta path-based top-k
similarity search in heterogeneous information networks. Proceedings of the VLDB Endowment 4(11),
992–1003.
Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and Jian Tang. 2019. RotatE: Knowledge Graph Embedding
by Relational Rotation in Complex Space. In 7th International Conference on Learning Representations,
ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net.
Table of Contents
Mihai Surdeanu, Julie Tibshirani, Ramesh Nallapati, and Christopher D. Manning. 2012. Multi-instance
Multi-label Learning for Relation Extraction. In Proceedings of the 2012 Joint Conference on Empirical
Methods in Natural Language Processing and Computational Natural Language Learning, EMNLP-CoNLL
2012, July 12-14, 2012, Jeju Island, Korea, Jun’ichi Tsujii, James Henderson, and Marius Pasça (Eds.).
The Association for Computational Linguistics, 455–465.
Zafar Habeeb Syed, Michael Röder, and Axel-Cyrille Ngonga Ngomo. 2018. FactCheck: Validating RDF
Triples Using Textual Evidence. In Proceedings of the 27th ACM International Conference on Information
and Knowledge Management, CIKM 2018, Torino, Italy, October 22-26, 2018, Alfredo Cuzzocrea, James
Allan, Norman W. Paton, Divesh Srivastava, Rakesh Agrawal, Andrei Z. Broder, Mohammed J. Zaki,
K. Selçuk Candan, Alexandros Labrinidis, Assaf Schuster, and Haixun Wang (Eds.). ACM Press, 1599–
1602.
Zafar Habeeb Syed, Michael Röder, and Axel-Cyrille Ngonga Ngomo. 2019. Unsupervised Discovery of
Corroborative Paths for Fact Validation. In The Semantic Web - ISWC 2019 - 18th International
Semantic Web Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part I, Chiara
Ghidini, Olaf Hartig, Maria Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song,
Maxime Lefrançois, and Fabien Gandon (Eds.). Lecture Notes in Computer Science, vol. 11778.
Springer, 630–646.
James Joseph Sylvester. 1878. Chemistry and Algebra. Nature 17, p284.
Jeremy Tandy, Ivan Herman, and Gregg Kellogg. 2015. Generating RDF from Tabular Data on the Web,
W3C Recommendation 17 December 2015. W3C Recommendation. World Wide Web Consortium.
December 17, 2015.
Jeni Tennison and Gregg Kellogg. 2015. Metadata Vocabulary for Tabular Data, W3C Recommendation 17
December 2015. W3C Recommendation. World Wide Web Consortium. December 17, 2015.
The R Foundation. 1992. The R Project for Statistical Computing. Online at https://www.r-project.org/.
The UniProt Consortium. 2014. UniProt: a hub for protein information. Nucleic Acids Research 43(D1),
D204–D212.
Bryan B. Thompson, Mike Personick, and Martyn Cutcher. 2014. The Bigdata® RDF Graph Database. In
Linked Data Management, Andreas Harth, Katja Hose, and Ralf Schenkel (Eds.). CRC Press, 193–237.
Katherine Thornton, Harold Solbrig, Gregory S. Stupp, José Emilio Labra Gayo, Daniel Mietchen, Eric
Prud’hommeaux, and Andra Waagmeester. 2019. Using Shape Expressions (ShEx) to Share RDF Data
Models and to Guide Curation with Rigorous Validation. In The Semantic Web - 16th International
Conference, ESWC 2019, Portorož, Slovenia, June 2-6, 2019, Proceedings, Pascal Hitzler, Miriam
Fernández, Krzysztof Janowicz, Amrapali Zaveri, Alasdair J. G. Gray, Vanessa López, Armin Haller,
and Karl Hammar (Eds.). Lecture Notes in Computer Science, vol. 11503. Springer, 606–620.
Felice Tobin. 2017. Thomson Reuters Launches first of its kind Knowledge Graph Feed allowing Financial
Services customers to accelerate their AI and Digital Strategies. Thomson Reuters Press Release.
October 23, 2017.
Dominik Tomaszuk, Renzo Angles, Lukasz Szeremeta, Karol Litman, and Diego Cisterna. 2019.
Serialization for Property Graphs. In Beyond Databases, Architectures and Structures. Paving the Road to
Smart Data Processing and Analysis - 15th International Conference, BDAS 2019, Ustroń, Poland, May
Give us feedback!
https://kgbook.org 137/150
2/15/23, 5:13 PM Knowledge Graphs
28-31, 2019, Proceedings, Stanislaw Kozielski, Dariusz Mrozek, Pawel Kasprowski, Bozena Malysiak-
Mrozek, and Daniel Kostrzewa (Eds.). Communications in Computer and Information Science,
vol. 1018. Springer, 57–69.
Gerald Töpper, Magnus Knuth, and Harald Sack. 2012. DBpedia ontology enrichment for inconsistency
detection. In I-SEMANTICS 2012 - 8th International Conference on Semantic Systems, I-SEMANTICS '12,
Graz, Austria, September 5-7, 2012, Valentina Presutti and Helena Sofia Pinto (Eds.). ACM Press, 33–
40.
Jeffrey Travers and Stanley Milgram. 1969. An Experimental Study of the Small World Problem.
Sociometry 32(4), 425–443.
Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier, and Guillaume Bouchard. 2016.
Complex Embeddings for Simple Link Prediction. In Proceedings of the 33nd International Conference
on Machine Learning, ICML 2016, New York City, NY, USA, June 19-24, 2016, Maria-Florina Balcan
and Kilian Q. Weinberger (Eds.). JMLR Workshop and Conference Proceedings, vol. 48. JMLR.org,
2071–2080.
Ledyard R. Tucker. 1964. The extension of factor analysis to three-dimensional matrices. In Contributions
to Mathematical Psychology, H. Gulliksen and N. Frederiksen (Eds.). Holt, Rinehart and Winston, 110–
Table of Contents
127.
Giovanni Tummarello, Christian Morbidoni, Reto Bachmann-Gmür, and Orri Erling. 2007. RDFSync:
Efficient Remote Synchronization of RDF Models. In The Semantic Web, 6th International Semantic
Web Conference, 2nd Asian Semantic Web Conference, ISWC 2007 + ASWC 2007, Busan, Korea,
November 11-15, 2007, Karl Aberer, Key-Sun Choi, Natasha Fridman Noy, Dean Allemang, Kyung-Il
Lee, Lyndon J. B. Nixon, Jennifer Golbeck, Peter Mika, Diana Maynard, Riichiro Mizoguchi, Guus
Schreiber, and Philippe Cudré-Mauroux (Eds.). Lecture Notes in Computer Science, vol. 4825.
Springer, 537–551.
Octavian Udrea, Diego Reforgiato Recupero, and V. S. Subrahmanian. 2010. Annotated RDF. ACM
Transactions on Computational Logics 11(2), 10:1–10:41.
Jacopo Urbani, Spyros Kotoulas, Jason Maassen, Frank van Harmelen, and Henri E. Bal. 2012. WebPIE: A
Web-scale Parallel Inference Engine using MapReduce. Journal of Web Semantics 10, 59–75.
Hernán Vargas, Carlos Buil Aranda, Aidan Hogan, and Claudia López. 2019. RDF Explorer: A Visual
SPARQL Query Builder. In The Semantic Web - ISWC 2019 - 18th International Semantic Web
Conference, Auckland, New Zealand, October 26-30, 2019, Proceedings, Part I, Chiara Ghidini, Olaf
Hartig, Maria Maleshkova, Vojtech Svátek, Isabel F. Cruz, Aidan Hogan, Jie Song, Maxime
Lefrançois, and Fabien Gandon (Eds.). Lecture Notes in Computer Science, vol. 11778. Springer,
647–663.
Shikhar Vashishth, Prince Jain, and Partha Talukdar. 2018. CESI: Canonicalizing Open Knowledge Bases
using Embeddings and Side Information. In Proceedings of the 2018 World Wide Web Conference on
World Wide Web, WWW 2018, Lyon, France, April 23-27, 2018, Pierre-Antoine Champin, Fabien L.
Gandon, Mounia Lalmas, and Panagiotis G. Ipeirotis (Eds.). ACM Press, 1317–1327.
Paola Velardi, Stefano Faralli, and Roberto Navigli. 2013. OntoLearn Reloaded: A Graph-Based Algorithm
for Taxonomy Induction. Computational Linguistics 39(3), 665–707.
Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio.
2018. Graph Attention Networks. In 6th International Conference on Learning Representations, ICLR
2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings. OpenReview.net.
12 pages.
Ruben Verborgh, Miel Vander Sande, Pieter Colpaert, Sam Coppens, Erik Mannens, and Rik Van de
Walle. 2014. Web-Scale Querying through Linked Data Fragments. In Proceedings of the Workshop on
Linked Data on the Web, co-located with the 23rd International World Wide Web Conference (WWW
2014), Seoul, Korea, April 8, 2014, Christian Bizer, Tom Heath, Sören Auer, and Tim Berners-Lee
(Eds.). CEUR Workshop Proceedings, vol. 1184. Sun SITE Central Europe (CEUR). 10 pages.
Ruben Verborgh, Miel Vander Sande, Olaf Hartig, Joachim Van Herwegen, Laurens De Vocht, Ben De
Meester, Gerald Haesendonck, and Pieter Colpaert. 2016. Triple Pattern Fragments: A low-cost
knowledge graph interface for the Web. Journal of Web Semantics 37–38, 184–206.
Serena Villata and Fabien Gandon. 2012. Licenses Compatibility and Composition in the Web of Data. In
Proceedings of the Third International Workshop on Consuming Linked Data, COLD 2012, Boston, MA,
USA, November 12, 2012, Juan F. Sequeda, Andreas Harth, and Olaf Hartig (Eds.). CEUR Workshop
Proceedings, vol. 905. Sun SITE Central Europe (CEUR). 12 pages. Give us feedback!
https://kgbook.org 138/150
2/15/23, 5:13 PM Knowledge Graphs
Serena Villata, Nicolas Delaforge, Fabien Gandon, and Amelie Gyrard. 2011. An Access Control Model for
Linked Data. In On the Move to Meaningful Internet Systems: OTM 2011 Workshops - Confederated
International Workshops and Posters: EI2N+NSF ICE, ICSP+INBAST, ISDE, ORM, OTMA,
SWWS+MONET+SeDeS, and VADER 2011, Hersonissos, Crete, Greece, October 17-21, 2011.
Proceedings, Robert Meersman, Tharam S. Dillon, and Pilar Herrero (Eds.). Lecture Notes in
Computer Science, vol. 7046. Springer, 454–463.
Johanna Völker, Daniel Fleischhacker, and Heiner Stuckenschmidt. 2015. Automatic Acquisition of Class
Disjointness. Journal of Web Semantics 35(P2), 124–139.
Julius Volz, Christian Bizer, Martin Gaedke, and Georgi Kobilarov. 2009. Discovering and Maintaining
Links on the Web of Data. In The Semantic Web - ISWC 2009, 8th International Semantic Web
Conference, ISWC 2009, Chantilly, VA, USA, October 25-29, 2009. Proceedings, Abraham Bernstein,
David R. Karger, Tom Heath, Lee Feigenbaum, Diana Maynard, Enrico Motta, and Krishnaprasad
Thirunarayan (Eds.). Lecture Notes in Computer Science, vol. 5823. Springer, 650–665.
Denny Vrandečić and Markus Krötzsch. 2014. Wikidata: A Free Collaborative Knowledgebase.
Communications of the ACM 57(10), 78–85.
Andreas Wagner, Duc Thanh Tran, Günter Ladwig, Andreas Harth, and Rudi Studer. 2012. Top-$k$
Table of Contents
Linked Data Query Processing. In The Semantic Web: Research and Applications - 9th Extended Semantic
Web Conference, ESWC 2012, Heraklion, Crete, Greece, May 27-31, 2012. Proceedings, Elena Simperl,
Philipp Cimiano, Axel Polleres, Óscar Corcho, and Valentina Presutti (Eds.). Lecture Notes in
Computer Science, vol. 7295. Springer, 56–71.
Claudia Wagner, Eduardo Graells-Garrido, David García, and Filippo Menczer. 2016. Women through the
glass ceiling: gender asymmetries in Wikipedia. EPJ Data Science 5(1), p5.
Zhen Wang, Jianwen Zhang, Jianlin Feng, and Zheng Chen. 2014. Knowledge Graph Embedding by
Translating on Hyperplanes. In Proceedings of the Twenty-Eighth AAAI Conference on Artificial
Intelligence, July 27 -31, 2014, Québec City, Québec, Canada, Carla E. Brodley and Peter Stone (Eds.).
AAAI Press, 1112–1119.
Quan Wang, Bin Wang, and Li Guo. 2015. Knowledge Base Completion Using Embeddings and Rules. In
Proceedings of the Twenty-Fourth International Joint Conference on Artificial Intelligence, IJCAI 2015,
Buenos Aires, Argentina, July 25-31, 2015, Qiang Yang and Michael J. Wooldridge (Eds.). IJCAI/AAAI,
1859–1866.
Quan Wang, Zhendong Mao, Bin Wang, and Li Guo. 2017. Knowledge Graph Embedding: A Survey of
Approaches and Applications. IEEE Transactions on Knowledge and Data Engineering 29(12), 2724–
2743.
Meng Wang, Ruijie Wang, Jun Liu, Yihe Chen, Lei Zhang, and Guilin Qi. 2018. Towards Empty Answers
in SPARQL: Approximating Querying with RDF Embedding. In The Semantic Web - ISWC 2018 - 17th
International Semantic Web Conference, Monterey, CA, USA, October 8-12, 2018, Proceedings, Part I,
Denny Vrandečić, Kalina Bontcheva, Mari Carmen Suárez-Figueroa, Valentina Presutti, Irene Celino,
Marta Sabou, Lucie-Aimée Kaffee, and Elena Simperl (Eds.). Lecture Notes in Computer Science,
vol. 11136. Springer, 513–529.
Xiao Wang, Houye Ji, Chuan Shi, Bai Wang, Yanfang Ye, Peng Cui, and Philip S. Yu. 2019.
Heterogeneous Graph Attention Network. In The World Wide Web Conference, WWW 2019, San
Francisco, CA, USA, May 13-17, 2019, Ling Liu, Ryen W. White, Amin Mantrach, Fabrizio Silvestri,
Julian J. McAuley, Ricardo Baeza-Yates, and Leila Zia (Eds.). ACM Press, 2022–2032.
Gerhard Weikum and Martin Theobald. 2010. From information to knowledge: harvesting entities and re‐
lationships from web sources. In Proceedings of the Twenty-Ninth ACM SIGMOD-SIGACT-SIGART
Symposium on Principles of Database Systems, PODS 2010, June 6-11, 2010, Indianapolis, Indiana, USA,
Jan Paredaens and Dirk Van Gucht (Eds.). ACM Press, 65–76.
Robert West, Evgeniy Gabrilovich, Kevin Murphy, Shaohua Sun, Rahul Gupta, and Dekang Lin. 2014.
Knowledge Base Completion via Search-Based Question Answering. In 23rd International World Wide
Web Conference, WWW '14, Seoul, Republic of Korea, April 7-11, 2014, Chin-Wan Chung, Andrei Z.
Broder, Kyuseok Shim, and Torsten Suel (Eds.). ACM Press, 515–526.
Mark D. Wilkinson, Michel Dumontier, IJsbrand Jan Aalbersberg, Gabrielle Appleton, Myles Axton, Arie
Baak, Niklas Blomberg, Jan-Willem Boiten, Luiz Bonino da Silva Santos, Philip E. Bourne, Jildau
Bouwman, Anthony J. Brookes, Tim Clark, Mercè Crosas, Ingrid Dillo, Olivier Dumon, Scott
Edmunds, Chris T. Evelo, Richard Finkers, Alejandra Gonzalez-Beltran, Alasdair J.G. Gray, Paul
Groth, Carole Goble, Jeffrey S. Grethe, Jaap Heringa, Peter A.C ’t Hoen, Rob Hooft, Tobias Kuhn,
Give us feedback!
https://kgbook.org 139/150
2/15/23, 5:13 PM Knowledge Graphs
Ruben Kok, Joost Kok, Scott J. Lusher, Maryann E. Martone, Albert Mons, Abel L. Packer, Bengt
Persson, Philippe Rocca-Serra, Marco Roos, Rene van Schaik, Susanna-Assunta Sansone, Erik
Schultes, Thierry Sengstag, Ted Slater, George Strawn, Morris A. Swertz, Mark Thompson, Johan van
der Lei, Erik van Mulligen, Jan Velterop, Andra Waagmeester, Peter Wittenburg, Katherine
Wolstencroft, Jun Zhao, and Barend Mons. 2016. The FAIR Guiding Principles for scientific data
management and stewardship. Scientific Data 3. 9 pages.
William A. Woods. 1975. What’s in a Link: Foundations for Semantic Networks. In Representation and
Understanding, Daniel G. Bobrow and Allan Collins (Eds.). Studies in Cognitive Science. Elsevier, 35–
82.
Gong-Qing Wu, Ying He, and Xuegang Hu. 2018. Entity Linking: An Issue to Extract Corresponding Entity
With Knowledge Base. IEEE Access 6, 6220–6231.
Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and Philip S. Yu. 2019. A
Comprehensive Survey on Graph Neural Networks. CoRR abs/1901.00596. 22 pages.
Marcin Wylot, Manfred Hauswirth, Philippe Cudré-Mauroux, and Sherif Sakr. 2018. RDF Data Storage
and Query Processing Schemes: A Survey. ACM Computing Surveys 51(4), 84:1–84:36.
Guohui Xiao, Diego Calvanese, Roman Kontchakov, Domenico Lembo, Antonella Poggi, Riccardo Rosati,
Table of Contents
and Michael Zakharyaschev. 2018. Ontology-Based Data Access: A Survey. In Proceedings of the
Twenty-Seventh International Joint Conference on Artificial Intelligence, IJCAI 2018, July 13-19, 2018,
Stockholm, Sweden, Jérôme Lang (Ed.). IJCAI/AAAI, 5511–5519.
Reynold S. Xin, Joseph E. Gonzalez, Michael J. Franklin, and Ion Stoica. 2013. GraphX: a resilient dis‐
tributed graph system on Spark. In First International Workshop on Graph Data Management
Experiences and Systems, GRADES 2013, co-loated with SIGMOD/PODS 2013, New York, NY, USA, June
24, 2013, Peter A. Boncz and Thomas Neumann (Eds.). CWI/ACM, 2:1–2:6.
Reynold S. Xin, Josh Rosen, Matei Zaharia, Michael J. Franklin, Scott Shenker, and Ion Stoica. 2013.
Shark: SQL and rich analytics at scale. In Proceedings of the ACM SIGMOD International Conference on
Management of Data, SIGMOD 2013, New York, NY, USA, June 22-27, 2013, Kenneth A. Ross, Divesh
Srivastava, and Dimitris Papadias (Eds.). ACM Press, 13–24.
Wei Xu, Raphael Hoffmann, Le Zhao, and Ralph Grishman. 2013. Filling Knowledge Base Gaps for
Distant Supervision of Relation Extraction. In Proceedings of the 51st Annual Meeting of the Association
for Computational Linguistics, ACL 2013, 4-9 August 2013, Sofia, Bulgaria, Volume 2: Short Papers. The
Association for Computational Linguistics, 665–670.
Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. 2019. How Powerful are Graph Neural
Networks?. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA,
USA, May 6-9, 2019. OpenReview.net. 17 pages.
Bishan Yang, Wen-tau Yih, Xiaodong He, Jianfeng Gao, and Li Deng. 2015. Embedding Entities and
Relations for Learning and Inference in Knowledge Bases. In 3rd International Conference on Learning
Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, Yoshua
Bengio and Yann LeCun (Eds.). 12 pages.
Fan Yang, Zhilin Yang, and William W. Cohen. 2017. Differentiable Learning of Logical Rules for
Knowledge Base Reasoning. In Advances in Neural Information Processing Systems 30: Annual
Conference on Neural Information Processing Systems 2017, 4–9 December 2017, Long Beach, CA, USA,
Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna M. Wallach, Rob Fergus, S. V. N.
Vishwanathan, and Roman Garnett (Eds.), 2319–2328.
Luwei Yang, Zhibo Xiao, Wen Jiang, Yi Wei, Yi Hu, and Hao Wang. 2020. Dynamic Heterogeneous Graph
Embedding Using Hierarchical Attentions. In Advances in Information Retrieval - 42nd European
Conference on IR Research, ECIR 2020, Lisbon, Portugal, April 14-17, 2020, Proceedings, Part II, Joemon
M. Jose, Emine Yilmaz, João Magalhães, Pablo Castells, Nicola Ferro, Mário J. Silva, and Flávio
Martins (Eds.). Lecture Notes in Computer Science, vol. 12036. Springer, 425–432.
Taha Yasseri, Robert Sumi, András Rung, András Kornai, and János Kertész. 2012. Dynamics of Conflicts
in Wikipedia. PLOS One 7(6). 12 pages.
Xiaoxin Yin, Jiawei Han, and Philip S. Yu. 2008. Truth discovery with multiple conflicting information
providers on the web. IEEE Transactions on Knowledge and Data Engineering 20(6), 796–808.
Dani Yogatama, Daniel Gillick, and Nevena Lazic. 2015. Embedding Methods for Fine Grained Entity
Type Classification. In Proceedings of the 53rd Annual Meeting of the Association for Computational
Linguistics and the 7th International Joint Conference on Natural Language Processing of the Asian
Give us feedback!
https://kgbook.org 140/150
2/15/23, 5:13 PM Knowledge Graphs
Federation of Natural Language Processing, ACL 2015, July 26-31, 2015, Beijing, China, Volume 2: Short
Papers. The Association for Computational Linguistics, 291–296.
Amrapali Zaveri, Anisa Rula, Andrea Maurino, Ricardo Pietrobon, Jens Lehmann, and Sören Auer. 2016.
Quality assessment for Linked Data: A Survey. Semantic Web Journal 7(1), 63–93.
Fuzheng Zhang, Nicholas Jing Yuan, Defu Lian, Xing Xie, and Wei-Ying Ma. 2016. Collaborative knowl‐
edge base embedding for recommender systems. In Proceedings of the 22nd ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, August 13-17, 2016,
Balaji Krishnapuram, Mohak Shah, Alexander J. Smola, Charu C. Aggarwal, Dou Shen, and Rajeev
Rastogi (Eds.). ACM Press, 353–362.
Weizhen Zhang, Han Cao, Fei Hao, Lu Yang, Muhib Ahmad, and Yifei Li. 2019. The Chinese Knowledge
Graph on Domain-Tourism. In Advanced Multimedia and Ubiquitous Engineering, MUE/FutureTech
2019, James J. Park, Laurence T. Yang, Young-Sik Jeong, and Fei Hao (Eds.). Lecture Notes in
Electrical Engineering, vol. 590. Springer, 20–27.
Lei Zhang. 2002. Knowledge Graph Theory and Structural Parsing. Ph.D. dissertation. University of Twente,
Enschede, Netherlands.
Mingbo Zhao, Tommy WS Chow, Zhao Zhang, and Bing Li. 2015. Automatic image annotation via com‐
Table of Contents
Appendix A
Background
We now discuss the broader historical context that has paved the way for the modern advent of
knowledge graphs, and the definitions of the notion of “knowledge graph” that have been proposed
both before and after the announcement of the Google Knowledge Graph [Singhal, 2012]. We re‐
mark that the discussion presented here builds upon (but does not subsume) previous discussion by
Ehrlinger and Wöß [2016] and Bergman [2019], which we refer to for further details. Though our
Give us feedback!
goal is to be comprehensive, the list of historical references should not be considered exhaustive.
https://kgbook.org 141/150
2/15/23, 5:13 PM Knowledge Graphs
The lineage of knowledge graphs can be traced back to the origins of diagrammatic forms of knowl‐
edge representation: a tradition going back at least as far as Aristotle (∼350 BC), followed by notions
such as Euler circles and Venn diagrams that helped humans to reason through visual insights.
Centuries later, a variety of researchers – particularly Sylvester [1878], Peirce [1878] and Frege
[1879] – independently devised formal diagrammatic systems that not only facilitate reasoning, but
also codify reasoning; in other words, their goal was to use diagrams as formal systems.
With the advent of digital computers, programs began to be used to perform formal reasoning and to
code representations of knowledge. These developments can be traced back to works such as those of
Richens [1958], Quillian [1963], and Travers and Milgram [1969], which focused on formal repre‐
sentations for natural language, information, and knowledge. These early works were limited (at
least by modern standards) by the poor computational resources available. From the formal (logical)
point of view, a number of influential developments took place in the 70's, including the introduc‐
tion of frames by Minsky [1974], the formalisation of semantic networks by Brachman [1977] and
Table of Contents
Woods [1975], and the proposal of conceptual graphs by Sowa [1979]. These works tried to integrate
formal logic with diagrammatic representations of knowledge by giving a (more-or-less) formal se‐
mantics to graph representations. But as Sowa [1979]a later wrote in the entry “Semantic networks”
of the Encyclopedia of Cognitive Science: “Woods (1975) and McDermott (1976) observed, the seman‐
tic networks themselves have no well-defined semantics. Standard predicate calculus does have a precisely
defined, model theoretic semantics; it is adequate for describing mathematical theories with a closed set of
axioms. But the real world is messy, incompletely explored, and full of unexpected surprises.”
From this era of exploration and attempts to define programs to simulate the visual and formal rea‐
soning of humans, the following key notions were established that are still of relevance today:
These works on conceptual graphs, semantic networks, and frames were direct predecessors of
Description Logics, which aimed to give a well-defined semantics to these earlier notions towards
building practical reasoning systems for decidable logics. Description Logics stem from the KL-ONE
system proposed by Brachman and Schmolze [1985], and the “attributive concept descriptions with
complements” language (aka ALC ) proposed by Schmidt-Schauß and Smolka [1991]. Description
Logics would be further explored in later years (see Section 4.3.2) and formed the underpinnings of
the Web Ontology Language (OWL) standard [Hitzler et al., 2012]. Together with the Resource
Description Framework (RDF) [Cyganiak et al., 2014], OWL would become one of the main building
blocks of the Semantic Web [Berners-Lee et al., 2001], within which many of the formative ideas
and standards underlying knowledge graphs would later be developed, including not only RDF and
OWL, but also RDFS [Brickley and Guha, 2014], SPARQL [Harris et al., 2013], Linked Data princi‐
ples [Berners-Lee, 2006], Shape Expressions [Brickley and Guha, 2014, Thornton et al., 2019], and
indeed, many of the other concepts, standards and techniques discussed in this book. Most of the
open knowledge graphs discussed in Section 10.1 – including BabelNet [Navigli and Ponzetto, 2012],
DBpedia [Lehmann et al., 2015], Freebase [Bollacker et al., 2007a], Wikidata [Vrandečić and
Krötzsch, 2014], YAGO [Suchanek et al., 2007], etc. – have either emerged from the Semantic Web
community, or would later adopt the standards it proposes.
Long before the 2012 announcement of the Google Knowledge Graph, various authors had used the
phrase “knowledge graph” in publications stretching back to the 40’s, but with unrelated meaning.
To the best of our knowledge, the first reference to a “knowledge graph” of relevance to the modern
meaning was in a paper by Schneider [1973] in the area of computerised instructional systems for
Give
education, where a knowledge graph – in his case a directed graph whose nodes are units of knowl‐ us feedback!
https://kgbook.org 142/150
2/15/23, 5:13 PM Knowledge Graphs
edge (concepts) that a student should acquire, and whose edges denote dependencies between such
units of knowledge – is used to represent and store an instructional course on a computer. An analo‐
gous notion of a “knowledge graph” was used by Marchi and Miguel [1974] to study paths through
the knowledge units of an instructional course that yield the highest payoffs for teachers and stu‐
dents in a game-theoretic sense. Around the same time, in a paper on linguistics, Kümmel [1973] de‐
scribes a numerical representation of knowledge, with “radicals” – referring to some symbol with
meaning – forming the nodes of a knowledge graph.
Further authors were to define instantiations of knowledge graphs in the 80’s. Rada [1986] defines a
knowledge graph in the context of medical expert systems, where domain knowledge is defined as a
weighted graph, over which a “gradual” learning process is applied to refine knowledge by making
small change to weights. Bakker [1987] defines a knowledge graph with the purpose of cumulatively
representing content gleaned from medical and sociological texts, with a focus on causal relation‐
ships. Work on knowledge graphs from the same group would continue over the years, with contri‐
butions by Stokman and de Vries [1988] further introducing mereological (part of) and instantiation
(is a) relations to the knowledge graph, and thereafter by James [1992], Hoede [1995], Zhang
Table of Contents
[2002], Popping [2003], amongst others, in the decades that followed [Nurdiati and Hoede, 2012].
The notion of knowledge graph used in such works considered a fixed number of relations. Other au‐
thors pursued their own parallel notions of knowledge graphs towards the end of the 80’s. Rappaport
and Gouyet [1988] describe a user interface for visualising a knowledge-base – composed of facts
and rules – using a knowledge graph that connects related elements of the knowledge-base. Srikanth
and Jarke [1989] use the notion of a knowledge graph to represent the entities and relations in‐
volved in projects, particularly software projects, where partitioning techniques are applied to the
knowledge graph to modularise the knowledge required in the project.
Continuing to the 90’s, the notion of a “knowledge graph” would again arise in different, seemingly
independent settings. De Raedt et al. [1990] propose a knowledge graph as a directed graph com‐
posed of a taxonomy of instances being related with weighted edges to a taxonomy of classes; they
use symbolic learning to extract such knowledge graphs from examples. Machado and Freitas da
Rocha [1990] define a knowledge graph as an acyclic, weighted and–or graph, 3939defining fuzzy de‐
note An and–or graph denotes dependency re
of sub-goals
pendencies that connect observations to hypotheses through intermediary nodes. These on which a goal depends, while or d
knowledge
graphs are elicited from domain experts and can be used to generate neural networks for selecting
hypotheses from input observations. Knowledge graphs were again later used by Dieng et al. [1992]
to represent the results of knowledge acquisition from experts. Shimony et al. [1997] rather define a
knowledge graph based on a Bayesian knowledge base – i.e., a Bayesian network that permits directed
cycles – over which Bayesian inference can be applied. This definition was further built upon in a
later work by Santos Jr. and Santos [1999].
Moving to the 00’s, Jiang and Ma [2002] introduce the notion of “plan knowledge graphs” where
nodes represent goals and edges dependencies between goals, further encoding supporting degrees
that can change upon further evidence. Search algorithms are then defined on the graph to deter‐
mine a plan for a particular goal. Helms and Buijsrogge [2005] propose a knowledge graph to repre‐
sent the flow of knowledge in an organisation, with nodes representing knowledge actors (creators,
sharers, users), edges representing knowledge flow from one actor to another, and edge weights indi‐
cating the “velocity” (delay of flow) and “viscosity” (the depth of knowledge transferred). Graph al‐
gorithms are then proposed to find bottlenecks in knowledge flow. Kasneci et al. [2008] propose a
search engine for knowledge graphs, defined to be weighted directed edge-labelled graphs, where
weights denote confidence scores based on the centrality of source documents from which the
edge/relation was extracted. From the same group, Elbassuoni et al. [2009] adopt a similar notion of
a knowledge graph, adding edge attributes to include keywords from the source, a count of support‐
ing sources, etc., showing how the graph can be queried. Coursey and Mihalcea [2009] construct a
knowledge graph from Wikipedia, where nodes represent Wikipedia articles and categories, while
edges represent the proximity of nodes. Given an input text, entity linking and centrality measures
are applied over the knowledge graph to determine relevant Wikipedia categories for the text.
Concluding with the 10’s (prior to 2012), Pechsiri and Piriyakul [2010] use knowledge graphs to
capture “explanation knowledge” – the knowledge of why something is the way it is – by represent‐
Give us feedback!
ing events as nodes and causal relationships as edges, claiming that this graphical notation offers
https://kgbook.org 143/150
2/15/23, 5:13 PM Knowledge Graphs
more intuitive explanations to users; their work focuses on extracting such graphs from text. Corby
and Faron-Zucker [2010] use the phrase “knowledge graph” in a general way to denote any graph
encoding knowledge, proposing an abstract machine for querying such graphs.
Other phrases were used to represent similar notions by other authors, including “information
graphs” [Kümmel, 1973], “information networks” [Sun et al., 2011], “knowledge
networks” [Ciampaglia et al., 2015], as well as “semantic networks” [Brachman, 1977, Woods, 1975,
Navigli and Ponzetto, 2012] and “conceptual graphs” [Sowa, 1979], as mentioned previously. Here
we exclusively considered works that (happen to) use the phrase “knowledge graph” prior to
Google’s announcement of their knowledge graph in 2012, where we see that many works had inde‐
pendently coined this phrase for different purposes. Similar to the current practice, all of the works
of this period consider a knowledge graph to be formed of a set of nodes denoting entities of interest
and a set of edges denoting relations between those entities, with different entities and relations be‐
ing considered in different works. Some works add extra elements to these knowledge graphs, such
as edge weights, edge labels, or other metadata [Elbassuoni et al., 2009]. Other trends include
knowledge acquisition from experts [Rada, 1986, Machado and Freitas da Rocha, 1990, Dieng et al.,
Table of Contents
1992] and knowledge extraction from text [Bakker, 1987, Stokman and de Vries, 1988, James, 1992,
Hoede, 1995], combinations of symbolic and inductive methods [Machado and Freitas da Rocha,
1990, De Raedt et al., 1990, Shimony et al., 1997, Santos Jr. and Santos, 1999], as well as the use of
rules [Rappaport and Gouyet, 1988], ontologies [Hoede, 1995], graph analytics [Srikanth and Jarke,
1989, Helms and Buijsrogge, 2005, Kasneci et al., 2008], learning [Rada, 1986, De Raedt et al.,
1990, Shimony et al., 1997, Santos Jr. and Santos, 1999], amongst other techniques. Later papers
(2008–2010) by Kasneci et al. [2008], Elbassuoni et al. [2009], Coursey and Mihalcea [2009] and
Corby and Faron-Zucker [2010] introduce notions of “knowledge graph” that are more similar to the
current practice.
However, some trends are not reflected in current practice. Of note is that many of the knowledge
graphs defined in this period consider edges as denoting a form of dependence or causality, where
x –➛ y may denote that x is a prerequisite for y [Schneider, 1973, Marchi and Miguel, 1974, Jiang
and Ma, 2002] or that x leads to y [Rada, 1986, Bakker, 1987, Rappaport and Gouyet, 1988,
Machado and Freitas da Rocha, 1990, Shimony et al., 1997, Jiang and Ma, 2002]. In some cases
and–or graphs are used to denote conjunctions or disjunctions of such relations [Machado and Freitas
da Rocha, 1990], while in other cases edges are weighted to assign a belief to a relation [Machado
and Freitas da Rocha, 1990, Jiang and Ma, 2002, Rada, 1986]. Papers from 1970–2000 tend to have
worked with small graphs, which contrasts with modern practice where knowledge graphs can reach
scales of millions or billions of nodes [Noy et al., 2019]: during this period, computational resources
were more limited [Schneider, 1973], and fewer sources of structured data were readily available,
meaning that the knowledge graphs were often sourced solely from human experts [Rada, 1986,
Machado and Freitas da Rocha, 1990, Dieng et al., 1992] or from text [Bakker, 1987, Stokman and
de Vries, 1988, James, 1992, Hoede, 1995].
Google Knowledge Graph was announced in 2012 [Singhal, 2012]. This initial announcement was
targeted at a broad audience, mainly motivating the knowledge graph and describing applications
that it would enable, where the knowledge graph itself is described as “[a graph] that understands
real-world entities and their relationships to one another” [Singhal, 2012]. Mentions of “knowledge
graphs” gained momentum in the research literature from that point. As noted by Bergman [2019],
this announcement by Google was a watershed moment for adopting the phrase “knowledge graph”.
However, given the informal nature of the announcement, a technical definition was
lacking [Ehrlinger and Wöß, 2016, Bonatti et al., 2018].
Given that knowledge graphs were gaining more and more attention not only in practice, but also in
the academic literature, formal definitions were becoming a necessity in order to precisely charac‐
terise what they were, how they were structured, how they could be used, etc., and more generally
Give us feedback!
https://kgbook.org 144/150
2/15/23, 5:13 PM Knowledge Graphs
to facilitate their study in a precise manner. We can determine four general categories of definitions
that have emerged.
Category I: The first category simply defines the knowledge graph as a graph where nodes represent
entities, and edges represent relationships between those entities. Often a directed edge-la‐
belled graph is assumed (or analogously, a set of named binary relations, or a set of triples).
This simple definition was popularised by seminal papers on knowledge graph
embeddings [Wang et al., 2014, Lin et al., 2015], being sufficient to represent the data model
upon which these embeddings operate. As reflected in the later survey by Wang et al. [2017],
the multitude of works that would follow on knowledge graph embeddings have continued to
use this definition. Though simple, the Category I definition raised some doubts: How is a
knowledge graph different from a graph (database)? Where does knowledge come into play?
Category II: A second common definition goes as follows: “a knowledge graph is a graph-structured
knowledge base”. To the best of our knowledge, the earliest usages of this definition in the liter‐
ature were by Nickel et al. [2016a] and Seufert et al. [2016] (interestingly in the formal nota‐
tion of these initial papers, a knowledge graph is defined analogously to a directed edge-la‐
Table of Contents
belled graph). Such a definition raises the question: what, then, is a “knowledge base”? The
phrase “knowledge base” was popularised in the 70’s (possibly earlier) in the context of rule-
based expert systems [Buchanan and Feigenbaum, 1978], and later were used in the context of
ontologies and other logical formalisms [Brachman and Schmolze, 1985]. The follow-up ques‐
tion then is: can we have a knowledge base (graph-structured or not) without a logical formal‐
ism while staying true to the original definitions? Looking in further detail, similar ambiguities
have also existed regarding the definition of a “knowledge base” (KB). Of note: Brachman and
Levesque [1986] – reporting after a workshop on this issue – state that “if we ask what the KB
tells us about the world, we are asking about its Knowledge Level”.
Category III: The third category of definitions outline more specific technical characteristics that a
“knowledge graph” should comply with.
In an influential survey on knowledge graph refinement, Paulheim [2017] lists four crite‐
ria that characterise the knowledge graphs considered for the paper. Specifically, he puts
forward that a knowledge graph “mainly describes real world entities and their interrelations,
organized in a graph; defines possible classes and relations of entities in a schema; allows for
potentially interrelating arbitrary entities with each other; covers various topical domains”; he
thus rules out ontologies without instances (e.g., DOLCE) and graphs of word senses (e.g.,
WordNet) as not meeting the first two criteria, while relational databases do not meet the
third criterion (due to schema restrictions), and domain-specific graphs (e.g., Geonames)
are considered to not meet the fourth criterion; this leaves graphs such as DBpedia,
YAGO, Freebase, etc.
Ehrlinger and Wöß [2016] also review definitions of “knowledge graph”, where they crit‐
icise the Category II definitions based on the argument that knowledge bases are often
synonymous with ontologies40, while knowledge graphs are not; they
note further
40 Prior criticiseof an ontology – such as
definitions
Google for calling its knowledge graph a “knowledge base”. Aftercontradict
reviewingthis conclusion.
prior defini‐
tions of terms such as “knowledge base”, “ontology”, and “knowledge graph”, they pro‐
pose their definition: “A knowledge graph acquires and integrates information into an ontol‐
ogy and applies a reasoner to derive new knowledge”. In the subsequent discussion, they re‐
mark that a knowledge graph is distinguished from an ontology (considered synonymous
with a knowledge base) by the provision of reasoning capabilities.
One of the most detailed technical definitions for a “knowledge graph” is provided by
Bellomarini et al. [2019], who state: “A knowledge graph is a semi-structured data model
characterized by three components: (i) a ground extensional component, that is, a set of rela‐
tional constructs for schema and data (which can be effectively modeled as graphs or general‐
izations thereof); (ii) an intensional component, that is, a set of inference rules over the con‐
structs of the ground extensional component; (iii) a derived extensional component that can be
produced as the result of the application of the inference rules over the ground extensional com‐
ponent (with the so-called “reasoning” process).” They remark that ontologies and rules rep‐
resent analogous structures, and that a knowledge graph is then a knowledge base ex‐
Give us feedback!
https://kgbook.org 145/150
2/15/23, 5:13 PM Knowledge Graphs
tended with reasoning along similar lines to the definition provided by Ehrlinger and
Wöß [2016].
We refer to Bergman [2019] for a list of further definitions that fit Category III. While having a
specific, technical definition for knowledge graphs provides a more solid foundation for their
study, as Bergman [2019] remarks, many such definitions do not seem to fit the current prac‐
tice of knowledge graphs. For example, it is not clear which of these definitions the Google
Knowledge Graph itself – responsible for popularising the idea – would meet (if any). Many of
the criteria proposed by such definitions are also orthogonal to the many works in the area of
knowledge graph embeddings [Wang et al., 2017].
Category IV: While the previous three categories involve (sometimes conflicting) intensional defini‐
tions, the fourth category adopts an extensional definition of knowledge graphs, defining them
in terms of prominent examples of knowledge graphs, such as DBpedia, Google’s Knowledge
Graph, Freebase, YAGO, amongst others [Bonatti et al., 2018]. Arguably this category avoids
the issue of defining a knowledge graph, rather than actually defining them.
These categories refer to definitions that have appeared in the academic literature. In terms of enter‐
Table of Contents
prise knowledge graphs, an important reference is the paper of Noy et al. [2019], which has been co-
authored by leaders of knowledge graph projects from eBay, Facebook, Google, IBM, and Microsoft,
and thus can be seen as representing a form of consensus amongst these companies – who have
played a key role in the popularisation of knowledge graphs – on what a “knowledge graph” means
in this setting. Specifically this paper states that “a knowledge graph describes objects of interest and
connections between them”, and goes on to state that “many practical implementations impose constraints
on the links in knowledge graphs by defining a schema or ontology”. They later add “Knowledge graphs
and similar structures usually provide a shared substrate of knowledge within an organization, allowing dif‐
ferent products and applications to use similar vocabulary and to reuse definitions and descriptions that
others create. Furthermore, they usually provide a compact formal representation that developers can use to
infer new facts and build up the knowledge”. We interpret this definition as corresponding to Category I,
but further acknowledging that while not a necessary condition for a knowledge graph, ontologies
and formal representations usually play a key role. The definition we provide at the outset of the pa‐
per is largely compatible with that of Noy et al. [2019].
Authors’ Biographies
Aidan Hogan
Aidan Hogan is an Associate Professor at the Department of Computer Science, Universidad de Chile,
where he also holds the position of Associate Researcher in the Millennium Institute for
Foundational Research on Data (IMFD). He received a B.Eng. and Ph.D. from the National University
of Ireland, Galway, in 2006 and 2011, respectively. His primary research interests centre on the
Semantic Web and Knowledge Graphs. He is the author of over one hundred research publications on
these topics, including two other books: “Reasoning Techniques for the Web of Data” and “The Web
of Data”.
Eva Blomqvist
Eva Blomqvist is an Associate Professor at the Department of Computer and Information Science,
Give
Linköping University. She received a Ph.D. from Linköping University, Sweden, in 2009, in the area us feedback!
https://kgbook.org 146/150
2/15/23, 5:13 PM Knowledge Graphs
of Ontology Learning for the Semantic Web. After a postdoc at ISTC-CNR in Rome, Italy, she has
been a member of the Semantic Web group at Linköping University since 2011. Her primary research
interests include the Semantic Web and Knowledge Graphs, more specifically the development and
use of ontologies as schemas for Knowledge Graphs. She is the author of over fifty research publica‐
tions in the area, and has served as scientific program chair of several of the top conferences in the
field.
Michael Cochez
Michael Cochez is an Assistant Professor in the Knowledge Representation and Reasoning Group at
the Computer Science department of the Vrije Universiteit, Amsterdam. He received his B.Sc. from
the University of Antwerp, Belgium and his M.Sc. and Ph.D. degrees from the University of
Jyväskylä, Finland. His research interests are in the intersection of Machine Learning and Knowledge
Graphs.
Table of Contents
Claudia d’Amato
Claudia d’Amato is an Associate Professor at the Department of Computer Science, University of Bari,
Italy and a member of the Knowledge Acquisition and Machine Learning Lab. She also holds a habili‐
tation as Full Professor for the scientific sectors: INF/01 and ING-INF/05. She received her Masters
Degree and Ph.D. from the University of Bari, Italy, in 2003 and 2007, respectively. Over the years,
she has also spent several invited-researcher stays in different international universities and research
institutes. Her primary research interests centre on Machine Learning for the Semantic Web and
Knowledge Graphs. She is the author of over one hundred research publications on these topics.
Gerard de Melo
Gerard de Melo is a Full Professor at the Hasso Plattner Institute for Digital Engineering and at the
University of Potsdam, where he holds the Chair for Artificial Intelligence and Intelligent Systems
and heads the corresponding research group. Previously, he was a faculty member at Rutgers
University in New Jersey and at Tsinghua University in Beijing, and a Post-Doctoral Research
Scholar at ICSI/UC Berkeley. He has published over 150 papers on natural language processing,
knowledge graphs, and AI, and received a number of best paper awards.
Claudio Gutierrez
Claudio Gutierrez is Full Professor at the Department of Computer Science, Universidad de Chile. He
is also a Senior Researcher in the Millennium Institute for Foundational Research on Data (IMFD).
His main research interests are the computational foundations of data and knowledge. He has
worked and published extensively in the areas of the Semantic Web and Databases, fields in which
he received test of time awards (ISWC and PODS). He also devotes time to research in the field of the
History of Science and Technology.
Sabrina Kirrane
Sabrina Kirrane is an Assistant Professor at the Vienna University of Economics and Business
Institute for Information Systems and New Media, where she is also a member of the Research
Institute for Cryptoeconomics and the Sustainable Computing Lab. Her research interests include
Security, Privacy, and Policy aspects of the Next Generation Internet (NGI), Distributed and
Decentralised Systems, Big Data and Data Science, with a particular focus on policy representation
and reasoning (e.g., access constraints, usage policies, regulatory obligations, societal norms, busi‐
ness processes), and the development of transparency and trust techniques for the Web. Give us feedback!
https://kgbook.org 147/150
2/15/23, 5:13 PM Knowledge Graphs
Jose Emilio Labra Gayo is an Associate Professor at the University of Oviedo, Spain. He founded the
WESO (Web Semantics Oviedo) research group in 2004, whose main goal is to apply semantic tech‐
nologies to solve practical problems. He was a member of the W3C Data Shapes working group and
is a member of the W3C Community Groups: Shape Expressions and SHACL. He is coauthor of the
“Validating RDF data” book and maintains the ShEx and SHACL library SHaclEX as well as the on‐
line tools RDFShape and Wikishape. Previously, he was coordinator of the Master in Web
Engineering and Dean of the School of Computer Science Engineering at the University of Oviedo
(2004–2012).
Roberto Navigli
Roberto Navigli is a Full Professor of Computer Science at the Sapienza University of Rome, where
he leads the Sapienza NLP Group. His research is focused on multilingual Natural Language
Table of Contents
Understanding, a field in which he received two grants of the European Research Council. In 2015
he received the META prize for groundbreaking work in overcoming language barriers with the
BabelNet lexical-semantic knowledge graph, a project also highlighted in The Guardian and Time
magazine, and winner of the Artificial Intelligence Journal prominent paper award 2017. He is the
co-founder of Babelscape, a successful company which enables Natural Language Understanding in
dozens of languages.
Sebastian Neumaier
Sebastian Neumaier is a researcher in the Data Intelligence group at the St. Poelten University of
Applied Sciences, Austria. He received an M.Sc. and Ph.D. from the Vienna University of Technology,
in 2015 and 2019, respectively. His Ph.D. thesis is centred around methods to facilitate the integra‐
tion and semantic enrichment of Open Data sources using Knowledge Graph technologies. His cur‐
rent research focuses on different aspects of semantic data management.
Axel-Cyrille Ngonga Ngomo is a Full Professor for Data Science at Paderborn University. He obtained
his M.Sc., Ph.D. and habilitation from the University of Leipzig, where he also led the Agile
Knowledge Engineering and Semantic Web Group. His research focuses on the automation of the life‐
cycle of knowledge graphs. Thus, his works include the development of approaches for the extrac‐
tion, integration, fusion, storage, analysis and exploitation of knowledge graphs.
Axel Polleres
Axel Polleres heads the Institute for Data, Process and Knowledge Management of Vienna University
of Economics and Business (WU Wien), which he joined in September 2013 as a Full Professor in the
area of “Data and Knowledge Engineering”. He is also a faculty member of the Complexity Science
Hub Vienna and was a visiting professor at Stanford University in 2018. He obtained his Ph.D. and
habilitation from Vienna University of Technology. His research focuses on ontologies, query lan‐
guages, logic programming, configuration technologies, Artificial Intelligence, Semantic Web, Linked
Open Data, Knowledge Graphs and their applications for Knowledge Management. Moreover, he ac‐
tively contributed to international standardisation efforts within the World Wide Web Consortium
(W3C) where he co-chaired the W3C SPARQL working group.
Sabbir M. Rashid
Give us feedback!
https://kgbook.org 148/150
2/15/23, 5:13 PM Knowledge Graphs
Sabbir M. Rashid is a Ph.D. candidate at Rensselaer Polytechnic Institute (RPI) working with
Deborah L. McGuinness on research related to data annotation and harmonisation, ontology engi‐
neering, knowledge representation, and various forms of reasoning. Prior to RPI, Sabbir completed a
double major at Worcester Polytechnic Institute, where he received B.Sc. degrees in both Physics and
Electrical & Computer Engineering. Much of his graduate studies at RPI have involved research relat‐
ed to data annotation and transformation using Semantic Data Dictionaries. His current work in‐
cludes the application of deductive and abductive inference techniques over Linked Health Data,
such as in the context of chronic diseases like diabetes.
Anisa Rula
knowledge technologies and data quality with a particular focus on data integration. She is research‐
ing new solutions to data integration with respect to the quality of data modelling and efficient solu‐
tions for large-scale data sources. Recently she has been working on data understanding for large and
complex datasets, on knowledge extraction, and on semantic data enrichment and refinement.
Lukas Schmelzeisen
Lukas Schmelzeisen is a PhD candidate working with Steffen Staab in the Analytic Computing group
at the University of Stuttgart, Germany. He holds a B.Sc. in Computer Science, which he received in
2015 at University of Koblenz–Landau. His main research interests are continuous representations of
both natural language corpora and knowledge graphs. In particular, his current focus is on how such
representations can be updated over time.
Juan Sequeda
Juan Sequeda is the Principal Scientist at data.world. He joined through the acquisition of Capsenta,
a company he founded as a spin-off from his research. His academic and industry work has been on
designing and building Knowledge Graph for enterprise data integration where he has researched
and developed technologies for semantic and graph data virtualisation, ontology and graph data
modelling and schema mapping, and data integration methodologies. Juan holds a Ph.D. in
Computer Science from the University of Texas at Austin. He is the recipient of the NSF Graduate
Research Fellowship, received 2nd Place in the 2013 Semantic Web Challenge for his work on
ConstituteProject.org, Best Student Research Paper at International Semantic Web Conference 2014
and the 2015 Best Transfer and Innovation Project awarded by the Institute for Applied Informatics.
Juan bridges academia and industry through standardisation committees, being a co-chair of the
Property Graph Schema Working Group, and past member of the Graph Query Languages task force
of the Linked Data Benchmark Council (LDBC), as well as a past invited expert member and stan‐
dards editor at the World Wide Web Consortium (W3C).
Steffen Staab
Steffen Staab holds a Cyber Valley endowed chair for Analytic Computing at the University of
Stuttgart, Germany, and a chair for Web and Computer Science at the University of Southampton,
UK. Steffen is a fellow of the European Association for Artificial Intelligence. His research interests
range from knowledge graphs and machine learning to the semantics of human--computer interac‐
tion. He is co-director of the Interchange Forum for Reflecting on Intelligent Systems (IRIS) at the
University of Stuttgart.
Give us feedback!
https://kgbook.org 149/150
2/15/23, 5:13 PM Knowledge Graphs
Antoine Zimmermann
☙ The End ❧
Table of Contents
Give us feedback!
https://kgbook.org 150/150