Accepted Manuscript

Semantic Web user interfaces – A model and a review

Charalampos C. Charalampidis, Euclid A. Keramopoulos

PII: S0169-023X(16)30208-7
DOI: 10.1016/j.datak.2018.04.003
Reference: DATAK 1642

To appear in: Data & Knowledge Engineering

Received Date: 22 September 2016

Revised Date: 4 March 2018
Accepted Date: 3 April 2018

Please cite this article as: C.C. Charalampidis, E.A. Keramopoulos, Semantic Web user interfaces – A
model and a review, Data & Knowledge Engineering (2018), doi: 10.1016/j.datak.2018.04.003.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Semantic Web User Interfaces – A Model and a Review

Charalampos C. Charalampidisa*, Euclid A. Keramopoulosa
a
Department of Information Technology, Alexander Technological Educational Institute of Thessaloniki, Greece

PT
Abstract
In the introduction of the Semantic Web vision, the software agents seek information, perform transactions and

RI
interact with physical devices. However, the Semantic Web is not yet fully implemented nor the software agents
are yet capable for this critical mission. The access of the Semantic Web is still a task mainly intended for the
humans. This access is through the user interfaces and is practiced mostly for information seeking tasks. The

SC
goal of this work is to create a review for the issues related to the user interfaces, with respect to their
application in the access of the Semantic Web. Therefore we build a model and a web application, to abstract the
interaction between the humans and the Semantic Web and investigate the features of the user interfaces as far
as the information seeking in the Semantic Web is concerned.

U
AN
At first a study of related literature is performed, and in it are identified and analyzed those distinctive
characteristics that a user interface needs to support. Then, it is conducted a field research in the World Wide
Web, in order to discover and record Semantic Web's user interfaces. Based on the analysis of the reviewed
literature, the model is devised, and the model's formalism is applied to the findings of the field research. After
M

that, it is conducted an evaluation study and with the help of a dedicated application, comparative tables are
outlined for reviewing user interfaces.
D

Keywords: SPARQL, Data structure, Data relation, Visualization, Exploration, Information Retrieval, Agents,
Zotero, EasyRDF
TE
C EP
AC

* Corresponding authors. E-mail addresses:

[email protected] (C. Charalampidis), [email protected] (E. Keramopoulos)
 ACCEPTED MANUSCRIPT

1. Introduction
The Semantic Web (SW) is conceived to be a network of interlinked structured and semi-structured data that are
semantically annotated. On the contrary, the World Wide Web (WWW) is a set of mainly unstructured data,
where the links (hyperlinks) to other data (documents) are just references. Moreover, the data structure in the
Semantic Web holds the information for the relation and the interconnection between them. This information is
modelled through the semantic annotation, applied with the RDF (Resource Definition Format) technology at

PT
the data interconnection level, while OWL (Web Ontology Language) provides the semantics for further
knowledge representation [1, 2, 3]. Both RDF and OWL formalisms are designed having in mind machines such
as computers or other modern intelligent digital devices, which will be able to interpret and understand the
information modelled in the semantically annotated data. The ultimate goal is to provide a web of interlinked

RI
structured data, where applications will be acting independently, on behalf of the humans and for the benefit of
the humans.

SC
The Semantic Web's technology is a stack of a number of other layered technologies such as data networking,
semantic annotation, knowledge representation, query of semantic data, reasoning and inferencing, software
user agents etc. The layer at a level of the stack is depended on the layer below it. This stack is presented in
Figure 1 [4]. In the full implementation of the Semantic Web, the humans will be interacting just with the

U
software user agents at the top of the semantic technologies stack. These agents will be doing the rest of the
work. However, currently, the upper layers, these of the user agents and the high level functions of reasoning
AN
and inferencing are not yet implemented.
M

Figure 1 (place holder)

File name “Figure1_layerCake-4.tiff”
D

scaled close to is original size

(width 63,50 mm, height 67,20 mm)
TE

Figure 1. The semantic Web stack

C EP
AC

In the original Scientific American article in May 2001, [1] the introduction of the Semantic Web's vision, is
presented through a hypothetical scenario where the software agents are retrieving information (“…The agent
promptly retrieved information about Mom's prescribed treatment from the doctor's agent, looked up several
lists of providers, and checked for the ones in-plan...”), are performing transactions (“…I'm going to have my
agent set up the appointments...”) and are interacting with physical devices (“…occurs when Pete answers his
phone and the stereo sound is turned down...”). Today the Semantic Web technology has achieved, in a
significant degree, the implementation of the information retrieval part. The majority of the Semantic Web
Applications (SWA) that include a user interface and can be easily located, are applications regarding
information retrieval. The other two parts of the original scenario (“transactions performance” and “interaction
with physical devices”) are likely to be integrated in some other applications [5] but it is rather difficult for the
user to know that it is about an application which integrates some of the Semantic Web technology (unless the
user is an insider, as far as the application is concerned).
 ACCEPTED MANUSCRIPT

For the human user, the Semantic Web technology, is meant (by conception and by design) to remain
inconceivable, invisible. The data in the Semantic Web are structured and linked/related. The structure and the
links are modelled in a way optimized for process by the machines. For the Human-Machine Interaction in the
context of the Semantic Web, the idea is to have the agents carry out the “assigned mission”. The human expects
from the agent the “mission accomplished”, marked with the delivery of the requested result. When the
Semantic Web is fully deployed, there will not be a special need for distinguishing a user interface as one for the
Semantic Web or for the World Wide Web. The user interfaces will be required in order the users can define the

PT
required tasks to software agents and get their response.

The requirement for a user interface specialized in accessing Semantic Web, appears when the user wants to
access the “raw” data of the Semantic Web (direct access of the Semantic Web by the humans). Such cases may

RI
be information retrieval applications or data processing applications or data mining applications [6]. The
challenge faced by a user interface to the Semantic Web is more complex than just getting information and
presenting documents or data. Furthermore, they have to take advantage of the RDF semantically annotated

SC
interlinked data [7], in order to present to the user useful information. On the other hand the user interface has to
hide the complexity of the Semantic Web's formalism in a way that is human understandable and user friendly.
The above extra features are required, beyond the well-known, general design principles and best practices. Of
course, these still do apply in their integrity.

Motivation, objectives and contribution:

U
AN
In practice, the Semantic Web access by the humans presents a twofold problem which in turn motivates our
work. The first fold concerns the need to interpret those special characteristics that a user interface has to
support, with respect to its application in the access of the Semantic Web. The second fold relates to the
M

adoption and use of new innovative types of data presentation in order to express the data structure and relations
inherent in the Semantic Web. These data structure and data relations constitute graphs (namely Semantic Web
graphs), and RDF/OWL triples representation as graphs, is a way to formalize but also visualize such structures
D

and relations.

The objective of this paper is to introduce an analysis model for the basic characteristics that a user interface
TE

applied to Semantic Web should support. Moreover, all the related user interfaces found in our review will be
evaluated based on our analysis model. Finally, the results of the critically review will be presented in a form of
comparative tables.
EP

The contribution of this paper is summarized as follows:

• An analysis model for the characteristics of the user interface, with respect to its application in the
access of the Semantic Web.
C

• The creation of comparative catalogue with Semantic Web applications, that are readily available for
AC

immediate interaction with the user. These applications were discovered through field research in the
World Wide Web, and the compilation of the catalogue is based on the analysis model mentioned
above.

• The delivery of the comparative catalogue through an interactive, Semantic Web application. In this
application, the human user is guided by the user interface to group or select the applications that are
best suited to his interest. As a result, the comparative catalogue is rather dynamic, since it may be
created on user’s demand and on the fly, based on the user’s knowledge and expertise in the Semantic
Web.

A key assumption: Since the Semantic Web is under development and its upper layers are still inadequately
implemented, we state the following assumption: When a user interacts with the Semantic Web, through any
 ACCEPTED MANUSCRIPT

kind of a user interface, this is done mostly to satisfy information needs [6] and less for processing existing data
or for adding new.

The rest of our work is organized as follows. In section 2 we study the literature about the diversification of the
user interfaces to the Semantic Web, with respect to those of the WWW. Then, in Section 3, we present the
development of our model for the analysis of the Semantic Web user interface. After that, in Section 4, we show
comparative tables for the reviewed user interfaces, produced from a special application we implemented [8],
[9]. Finally, we conclude this paper in Section 5, where we discuss the results of this survey and draw our

PT
conclusions and in Section 6, where we outline possible future work.

2. Prior work

RI
In this section, we introduce basic characteristics of Semantic Web Applications and their user interfaces, as
presented in prior woks. In the first part of the current section are presented prior works related to the user

SC
interface of SWA. In the second part of the current section are discussed issues related to information seeking in
the Semantic Web. Each of the presented prior works is addressing individual aspects of the interaction between
the humans and the Semantic Web. An approach which integrates these individual aspects in a single model, is
proposed by the authors of this work in the next section (section 3).

U
2.1 Semantic Web applications and their User Interfaces
AN
[10] introduces a model for the categorization of applications who are accessing the Semantic Web. The
categorization examines the conceptualization, design and implementation of a Semantic Web Application
(SWA). The categorization is based on the following characteristics.
M

• Semantic Technology Depth: The type of an SWA by the degree of the Semantic Web technology
utilization. According to this a SWA may be distinguished in “extrinsic” and “intrinsic”.
D

• Information Flow Direction: Whether the data are flowing from or to the application, in a consuming or
producing fashion.
TE

• Richness of Knowledge Representation: How rich is the SWA concerning the use of the knowledge
representation. This may be “shallow” or “strong”.

•
EP

Semantic Integration: The degree to what the reusable entities of the Semantic Web technology, are
integrated in the application. Whether the SWA in mind is semantically “Isolated” or semantically
“Integrated”.

•
C

User Involvement: The user's involvement may be considered through various aspects such as the
expertise in the Semantic Web technology, their contribution in the production, editing and refining the
AC

knowledge representation etc.

For the “Semantic Integration”, the authors of the work herein, would like to comment that all though the
“Semantic Integration” is a matter of choice or decision, it is also a matter of time and maturity. Over the years,
theories and tools grow, mature and establish their acceptance. Furthermore, the W3C encourages the integration
with the Semantic Web, with the prompt to apply reusable semantic entities, such as common RDF vocabularies
and ontologies. [11]

[12] performs a survey on the applications used for the visualization of the linked data and there are identified,
the requirements, features and design principles that should be supported by such applications. Also the end
users, as a key factor of the design, are categorized in the following three categories:

• Lay-users: They know how to use a computer and they are in position to seek information in the
internet, without necessarily understand the Semantic Web technology..
 ACCEPTED MANUSCRIPT

• Tech-users: Users who understand the Semantic Web and are familiar with the related technologies.

• Domain expert: They may not necessarily be experts in the Semantic Web technology but they can be
very efficient in using specialized tools for retrieving and interpreting data in their field of interest. In
[12] the non-SW domain experts are considered to be lay-users of Linked Data.

In [13] is presented an approach for the integrated visualization, exploration and authoring of semantically
enriched unstructured content. This approach is inspired by the WYSIWYM metaphor (What You See Is What

PT
You Mean). The term WYSIWYM describes the integrated visualization, exploration and authoring of
unstructured and semantic content [13]. With the use of a table, the WYSIWYM concept models the bindings
between semantic representation techniques on the one hand and user interface elements for visualizing,

RI
exploration and authoring on the other. Usually, the semantic representation techniques are implemented through
Trees, Graphs and Hypergraphs. Thus, in [13] the top level elements of a semantic representation (like Trees,
Graphs and Hypergraphs) are assigned to a group of columns in a table. In the same table, the top level user

SC
interface elements for the visualization, exploration and authoring, are assigned to a group of rows. The crossing
of a column and of a row of the table defines a binding between an element of semantic representation and an
element of the user interface (for the visualization, exploration and authoring).

In [14] are followed the findings and propositions of the related bibliography for the design of user interfaces

U
that are aesthetically pleasing. In particular effort is made to:
AN
• Develop a user interface for linked data.

• Support the human in the analysis of data, through visualization.

•
M

Create a user interface that is aesthetically pleasing, for use in the interactive navigation and
exploration

Based on all the above, in [14] was designed an application for the exploration of linked data where the
D

visualization is implemented through graphs. The transition from the design principles to the real visual design
considered the “Consistency in Visual Representations”, “Representing Semantic Concepts”, “Representing
TE

Semantic Relations”, “Layout of the nodelink graph” and “Interactions in the visualisation”.

In [15] a survey is conducted on existing semantic search engines in order to bring forward the features of the
interface design. This survey differs from other surveys about search engines, in the sense that it focuses on the
EP

features and elements of the user interface, rather than the relevance of the results.. In the reviewed search
engines were identified the elements and features used in their interfaces. The goal was to analyze and classify
the prevailing elements, features and approaches. Such are “Query Suggestion”, “Highlighted keyword/phrase”,
C

“Semantic Graph”, “Result Summary / Answer”, “Snippets View” etc. Having in mind that there are not any
specialized design guide-lines for user interfaces of the semantic search engines, the author suggests [15] some
AC

design guide-lines based on observations of the survey data. These suggestions are grouped in three groups
“General design guide-lines”, “Main Search Page design guide-lines”, “Search Engine Result Page design
guide-lines”.

2.2 Information seeking in the Semantic Web

In the World Wide Web, the information seeking tasks are performed over unstructured data and the expected
result is the delivery of a list of documents (the documents are considered in a wide sense: html, multimedia
files, pdf… etc.) [16] . The lists of the returned documents are usually ranked by their relevance to the searched
tokens and this relevance is judged by the intelligence embedded in the system by its designer.

On the other hand, in the Semantic Web the information seeking tasks are performed over structured data [17]
and the expected result is in the form of RDF triples as a response to a SPARQL query. Furthermore, the SWA
has to handle the presentation of RDF triples but also documents and even metadata . The knowledge of the
 ACCEPTED MANUSCRIPT

structure and/or relations of the retrieved data is of major importance or even mandatory for the interpretation of
the retrieved data. This knowledge of the structure is also likely to be the sole purpose of the information
seeking task. The relevance of the semantically annotated data to the user's information needs may not always
be apparent to the user, but it can be inferred through the processing of the structure and the relations metadata
[17] . In the context of the application, after the process of inference and interpretation of the meaning, the user
interface is responsible for the final presentation of the results in a human readable/user friendly manner [18].

In the Semantic Web, a user interface designed for information seeking, has to accomplish the following tasks:

PT
• get the user input for the search terms/phrase,

• display a list of ranked documents,

RI
• present the data, the structure and the RDF links, in a way that helps users to submit new queries based
on these results [19, 20] .

SC
3. Model for Analysing Semantic Web User Interfaces

U
The prior works presented in section 2 are addressing individual aspects of the interaction between the
humans and the Semantic Web. Following is presented an approach which integrates these individual
AN
aspects in a single model. Through this integration: a) Is brought forward the differentiation of the
returned data in the Semantic Web and its impact in the behaviour of the user interface, compared to
these of the World Wide Web. b) Are pointed out specific features of this differentiation. c) Is defined
M

a semantic context for the analysis of the Semantic Web User Interface.

3.1 The special characteristics of the user interface

D

A user interface for a SWA has to support some special characteristics in excess of the standards [3] , as
TE

described in theory for the graphical environment and its features related to usability [21] . Features such as tree
views, pull down menus, highlighting, text formatting, screen layout etc, are not special to the Semantic Web
and are not subject of this paper. These are already covered in other works and among them the [13] and [14] ,
and especially for the search engines, also in the [15] .
EP

Thus, based in Section 2, the special characteristics of user interfaces used in Semantic Web, are found to
converge to the following dimensions: “Users”, “Visual representation of the data and features of
navigation/exploration”, “Assistance to the user for the interpretation of the data”. These characteristics are
C

further analyzed/discussed below, as components of an integrated approached:

AC

• Users [10, 12] : The users are distinguished in “Lay-users”, “Tech-users” and “domain experts”.

For example, a reader of this paper is more likely to be a tech-user, while the user of Semantic Web
search engine, might be of any kind. On the other hand, any kind of user might need the service of a
search engine. In the same manner the users of a SPARQL query form, might be either tech-users or
even domain-users.

• Visual representation of the data and features of navigation/exploration [12, 13, 14, 22] :

What is visually represented is largely depended on what is returned as a response to a request. It may
be a list of documents (html, multimedia files, pdf…), a graph of an OWL ontology, an expanding tree
view guiding the browsing etc. The features of navigation/exploration are key characteristic which
complement the characteristic of the visual representation. Especially when there is the need for
successive seeks in the same session, the interactive facilities of navigation/exploration are necessary
 ACCEPTED MANUSCRIPT

[2] for the fast building of the queries. (i.e. in the combination of a Semantic Web browser and of a
query builder).

• Assistance to the user for the interpretation of the data:

In the case of a search engine in the Semantic Web [12, 17] , where the response to a user’s query is a
list of relevant documents, there is little need for a further interpretation by the user's reasoning, since
the mechanism of the interpretation is already embedded in the machine. However, in the case where

PT
the metadata are also returned (structure and/or relations) there is also the question who will perform
the task of the interpretation. If the machine is not capable for such a task, then it is the user who must
perform it, and he must have the knowledge for this.

RI
3.2 Setting up the model for the analysis of the Semantic Web User Interface
We organized and developed a field (internet) research, in order to find and review web applications, which are

SC
designed and implemented based on semantic web data and also consist of a graphical user interface. We found
thirty seven (37) such web applications. The full list is presented in Table 1.

In order to analyse all the reviewed web applications, we define a model whose dimensions are the special

U
characteristics that such an application has to support. A key factor in the design of the model, is the
differentiation of the returned data in the Semantic Web and its impact in the behavior of the user interface,
AN
compared to these of the World Wide Web. The Semantic Web's data are structured and the user interface needs
to represent their relations in a way that helps to interpret their meaning. Our intention for the definition of the
model is not to bypass the significance of the practices and methodologies of the graphical environment, but to
expose the differentiation which should be accounted additionally, when the access of the Semantic Web is
M

considered.

Thus, the model will focus on those specific features of this differentiation. More specifically, it is very
D

important for such a user interface to:

1. provide access to the data structure and their relation

TE

2. provide a special interface for the interpretation of the data relations.

In other words, the model will be defined so that it helps to distinguish those user interfaces that are closer to
EP

the human language and the human way of reasoning.

In the introduction was made the assumption that when a user of an application interacts with the Semantic Web
through any kind of a user interface, this is done mostly to satisfy information needs about the data and their
C

relations. Also, in subsection 3.1 was pointed out that the special characteristics of the user interfaces used in the
Semantic Web converge to the following dimensions:
AC

• Users (knowledge of the Semantic Web technology, computer skills)

• Visual representation of the data, structure and relations

• Features of navigation/exploration

• Interpretation of the data by the user

Since these dimensions do not describe the features that are special to the information seeking tasks [15] extra
dimensions will be included to cover this need.

The dimensions/attributes (features, characteristics) of the model are presented in the following paragraphs.
These dimensions/attributes are derived from the analysis presented above. Together with the attributes are
given the values these attributes may be assigned with. The values of all the attributes are nominal. As an
 ACCEPTED MANUSCRIPT

exception handling measure, any attribute may be assigned the value “undefined” (meaning “something else and
unknown”) if for any reason, is impossible to assign any other of the prescribed values. In the parenthesis is
noted an abbreviated form of these values. In square brackets is noted the term used in practice for the semantic
annotation in an RDF file (see Section 4). As “interactive” feature of the user interface, is deemed any feature
that can be triggered by the current action of the user, and once triggered so, its manifestation contributes to the
decision for the next action of the user.

The notation “SWF” for the “Output type” and “Input mode” attributes, stands for the phrase “Semantic Web

PT
Formalism”.

Type of User: Will be characterizing the user, with respect to his knowledge of the Semantic Web
technology and thus his ability to interpret data that are presented through the formalism of the

RI
Semantic Web. It is reminded that users of the “Domain expert” kind, that are not “Semantic Web
domain-experts”, are considered to be Lay-users. The possible values of this attribute will be:

SC
• Lay-user (Lay-user) [user_l]

• Tech-user (Tech-user) [user_t]

U
undefined (undefined) [user_u]

Information need of the user: This is implied from what can be searched for, in the Semantic web and
AN
hence what may be delivered as a response to an information need: “What can somebody search/find in
the Semantic Web and what is its difference with respect to the World Wide Web”. We may consider to a
significant extend, that the simple browsing or exploration in the Semantic Web, is an information
M

seeking task of the respective type (browsing or exploratory) [6, 14, 22]. This of course does not
exclude the other types of information seeking, namely (a) Direct search (known item search), (b)
Information look up (for an answer to a particular question) and (c) Exhaustive search (for anything
D

that is relative to the subject of interest). The information seeking which is different than the browsing
or the exploratory type, requires the support of additional user interface features. These features are
approached in a following paragraph regarding the input mode. The model's approach to the
TE

information need of the user does not deepen into the individual types of the information seeking.
Instead it will be limited to the differentiation of the information needs and to the differentiation of the
information seeking task, in the context of the Semantic Web. No matter what the information seeking
EP

type is, the focus is on what we are seeking (data, relations between the data or both data and their
relations)

The possible values of this attribute (Information need of the user) will be:
C

• Data (Data) [need_d]

AC

• Data structure (Structure) [need_s]

• Data and Data structure (Data and Structure) [need_a]

• undefined (undefined) [need_u]

Output type: What is the output returned? The “output type” will be the fusion of two elements,
namely, (a) the “Visual representation of the returned data, structure and relations” and (b) the
“Features of navigation/exploration”. This dimension combines the usability of human-machine
interaction with the need for “Interpretation of the data by the user” in the sense of the Semantic Web's
formalism. As the user traverses the hierarchy of the semantic web fields which are displayed in his
interface, he needs to know their meaning, in order to decide the next step in navigation/exploration.
Expressed in a manner closer to the human language, this could be relayed as the question of “how can
 ACCEPTED MANUSCRIPT

the user understand the meaning of what the machine returns and what can he do with it?” How can
the human understand and be benefited from the semantic data presented in the user interface?”
The possible values of this attribute will be :

• Documents such as html, multimedia files, pdf… etc. (Documents) [ret_d]

• With Semantic Web Formalism – text (with SWF – Text) [ret_swf_t]

• With Semantic Web Formalism – graphics (with SWF – Graphics) [ret_swf_g]

PT
• With Semantic Web Formalism – text and graphics (with SWF – Text & Graphics) [ret_swf_a]

• Without Semantic Web Formalism – text (no SWF – Text) [ret_nswf_t]

RI
• Without Semantic Web Formalism – graphics (no SWF – Graphics) [ret_nswf_g]

• Without Semantic Web Formalism – text and graphics (no SWF – Text & Graphics) [ret_nswf_a]

SC
• undefined (undefined) [ret_u]

The “text” notation in the above values, includes anything that is not presented as “graphic”.

U

Input mode: What are the elements of the user interface used to submit the information need? This
dimension is an abstract classification of such user interface elements, thus modeling the human-
AN
machine interaction, in the context of an information seeking task. In this kind of interaction, the user
communicates his information need to the underlying application, through user interface elements such
as query forms or query builders. Expressed in a manner closer to the human language, this could be
M

relayed as the question of “How can the user tell to the machine what he means or what he needs?”
The possible values of this attribute will be:

• Search term form, both simple or advanced (Search term form) [in_t_form]
D

• Interactive information seeking, with Semantic Web formalism (with SWF – interactive)
TE

[in_swf_int]

• Interactive information seeking, without Semantic Web formalism (no SWF – interactive)
[in_nswf_int]
EP

• SPARQL query form (SPARQL query form) [in_s_form]

• undefined (undefined) [in_u]

C

Application category: A way to characterize the application, based on both the kind of the user
(lay/tech – user) and the concept of his information need addressing. The possible values of this
AC

attribute will be:

• Search engine, with general or specialized content (Search engine) [cat_se]

• For Browsing and/or Exploration (Browser) [cat_br]

• SPARQL query builder, assistant or wizard (SPARQL query builder) [cat_qb]

• A tool for specialized users or a Domain-specific application (Tool) [cat_t]

• undefined (undefined) [cat_u]

Access mode: How a user can access the Semantic Web application? It examines if the application is
available just online or for download. The possible values of this attribute will be:
 ACCEPTED MANUSCRIPT

• Only on the WWW. One can use the application through the WWW, but just in the information
system of its maker (Online) [acc_on]

• Only local. One can use the application after installing it on a private information system (Local)
[acc_loc]

• On the WWW or Local. Both the two previous cases are possible (Online & Local) [acc_all]

• undefined (undefined) [acc_u]

PT
The formalism of the model described above, is applied on reviewed SWAs, resulting in Table 1 which is the
dataset to process further. The rows of the table correspond to the individual SWAs. The columns of the table

RI
correspond to the dimensions of the model (User, Information need of the user, Output type, Input mode,
Application category, Access mode). The attributes values of each single element of the table, were assigned
based on the formalism of the model and on the personal study of each element (semantic web application).

SC
Three extra columns were added to this table, for easier comprehension:

• Column “Site URL”; the application site whose user interface we study.

• Column “Title of the site”.

U
• Column “Comment” with some extra information for the site (if needed).
AN
Table 1. The data set. It is derived from the raw data collected with the field research, after they were treated
with the model's formalism. (“SWF” in the columns “Output type” and “Input mode” stands for “Semantic
M

Web Formalism”)
Information
Output Application Access
Site URL Title of the site User need of the Input mode Comment
D

type category mode

user
TE

http://datahub.io/ "Welcome - the Tech- with SWF – Search term A collection

1 Structure Browser Online
[23] Datahub" user Text form of datasets

"LodLive - A framework
http://en.lodlive.it Tech- Data and with SWF – with SWF – Online
2 browsing the Browser for visualizing
EP

/ [24] user Structure Graphics interactive & Local

Web of Data" linked data

"Encyclopedia
of Life - Domain-
C

http://eol.org/ Lay- no SWF – Search term Search

3 Animals - Data Online specific search
[25] user Text form engine
Plants - Pictures engine
AC

& Information"

http://graphite.ecs
"Q&D RDF Tech- with SWF – Search term
4 .soton.ac.uk/brow Structure Browser Online
Browser" user Text form
ser/ [26]

"LodView —
http://lodview.it/ Tech- with SWF – with SWF – Online
5 giving data a Structure Browser
27] user Graphics interactive & Local
new shape"
 ACCEPTED MANUSCRIPT

Information
Output Application Access
Site URL Title of the site User need of the Input mode Comment
type category mode
user

Plug-in for
web browsers
(may be tried
with
"OpenLink “http://www.b

PT
http://ode.openlin Tech- with SWF –
6 Data Explorer Structure undefined Tool Online bc.co.uk/musi
ksw.com/ [28] user Text
Extension" c/artists/ed2ac
1e9-d51d-

RI
4eff-a2c2-
85e81abd6360
”)

SC
"Swoogle with SWF –
http://swoogle.um Tech- Data and Search term Search
7 Semantic Web Text & Online
bc.edu/ [29] user Structure form engine
Search Engine" Graphics

U
"Virtuoso Member of
http://virit.it.teith Tech- Data and with SWF – SPARQL
8 SPARQL Query Tool Online the Virtuoso
e.gr/sparql/ [30] user Structure Text query form
AN
Editor" server suit

http://ws.nju.edu.
with SWF –
cn/falcons/objects "Falcons Object Tech- Data and Search term
9 Text & Browser Online
earch/index.jsp Search" user Structure form
M

Graphics
[31]

http://www.bing.c Lay- Search term Search

10 "Bing" Data Documents Online
om/ [32] user form engine
D

Displays also
http://www.cluuz. Lay- Search term Search
TE

11 "Cluuz Search" Data Documents Online a graph with

com/ [33] user form engine
related entities

PHP
framework, to
EP

"EasyRdf -
http://www.easyr Tech- consume and
12 RDF Library undefined undefined undefined Tool Online
df.org/ [34] user produce RDF
for PHP"
data with PHP
applications
C

http://www.exalea
"Web Search - Lay- Search term Search
AC

13 d.com/search/ Data Documents Online

Exalead" user form engine
[35]

"Factbites:
http://www.factbit Lay- Search term Search
14 Where results Data Documents Online
es.com/ [36] user form engine
make sense"

http://www.irisa.f
r/LIS/ferre/sparkli Lay- no SWF – no SWF – SPARQL
15 "Sparklis" Data Online
s/osparklis.html user Text interactive query builder
[37]

http://www.kngin "Kngine - Ask Lay- Search term Search

16 Data Documents Online
e.com/ [38] me anything" user form engine
 ACCEPTED MANUSCRIPT

Information
Output Application Access
Site URL Title of the site User need of the Input mode Comment
type category mode
user

http://www.ldodd "Twinkle: A Java based

Tech- SPARQL
17 s.com/projects/tw SPARQL Query undefined undefined Tool Local SPARQL
user query form
inkle/ [39] Tool" query form

"SenseBot -

PT
semantic search
http://www.sense Lay- Search term Search
18 engine that Data Documents Online
bot.net/ [40] user form engine
finds sense on
the Web"

RI
https://duckduckg Lay- Search term Search
19 "DuckDuckGo" Data Documents Online
o.com/ [41] user form engine

SC
https://ixquick.co "Ixquick Web Lay- Search term Search
20 Data Documents Online
m/ [42] Search" user form engine

Community-

U
curated
https://www.freeb Lay- Data and no SWF – Search term database of
21 "Freebase" Browser Online
AN
ase.com/ [43] user Structure Text form well-known
people, places,
and things

https://www.start "Startpage Web Lay- Search term Search

M

22 Data Documents Online

page.com/ [44] Search" user form engine

Free
D

knowledge
https://www.wiki
base with over
data.org/wiki/Wik Lay- Data and no SWF – Search term Search
23 "Wikidata" Online 13 million
TE

idata:Main_Page user Structure Text form engine

data items,
[45]
that anyone
can edit
EP

Domain-
http://www.dotac. "Tim Berners-
Lay- no SWF – no SWF – specific,
24 info/explorer/ Lee :: dotAC Data Browser Online
user Text interactive faceted search
[46] Explorer"
engine
C

Member of a
family of
AC

http://www.visual "Tools - Visual

Tech- with SWF – with SWF – applications
25 dataweb.org/gfac Data Web - Structure Browser Online
user Graphics interactive (http://www.vi
et/gfacet.php [47] gFacet"
sualdataweb.o
rg/tools.php)

Member of a
http://www.visual family of
"Tools - Visual
dataweb.org/relfi Lay- Data and no SWF – no SWF – applications
26 Data Web - Browser Online
nder/relfinder.php user Structure Graphics interactive (http://www.vi
RelFinder"
[48] sualdataweb.o
rg/tools.php)
 ACCEPTED MANUSCRIPT

Information
Output Application Access
Site URL Title of the site User need of the Input mode Comment
type category mode
user

Member of a
http://www.visual family of
"Tools - Visual
dataweb.org/seml Tech- with SWF – with SWF – applications
27 Data Web - Structure Browser Online
ens/semlens.php user Graphics interactive (http://www.vi
SemLens"
[49] sualdataweb.o

PT
rg/tools.php)

Member of a
family of

RI
http://www.visual "Tools - Visual
Tech- with SWF – with SWF – applications
28 dataweb.org/tface Data Web - Structure Browser Online
user Graphics interactive (http://www.vi
t/tfacet.php [50] tFacet"
sualdataweb.o

SC
rg/tools.php)

with SWF – Member of

http://virit.it.teith “OpenLink Tech- Data and with SWF – SPARQL
29 Text & Online the Virtuoso
e.gr/isparql/ [51] iSPARQL” user Structure interactive query builder

U
Graphics server suit

http://ws.nju.edu. “SView Tech- with SWF – Search term Search

AN
30 Structure Online
cn/sview/ [52] Homepage” user Text form engine

Provides
:: LED :: suggestions if
http://sisinflab.pol Lay- Search term Search
M

31 Lookup Explore Data Documents Online there are no

iba.it/led/ [53] user form engine
Discover query-relevant
data
D

Exploration of
http://code.know- CODE Linked with SWF – Linked Data
Tech- Data and Search term
TE

32 center.tugraz.at/se Data Query Text & Browser Online with tabular

user Structure form
arch [54] Wizard Graphics and graphic
analysis

Focuses on
EP

delivering
http://discoveryhu Discovery Hub Lay- no SWF – Search other relative
33 Data Documents Online
b.co/ [55] Beta user interactive engine data that
might be
C

interesting
AC

no SWF –
http://www.geona Lay- Search term
34 GeoNames Data Text & Tool Online *
mes.org/ [56] user form
Graphics

Swotti Opinion
http://www.swotti Analyser. Lay- Search term
35 Data Documents Tool Online **
.com/ [57] Search, Rate user form
and Compare

Transcript
https://linkedjazz. Analyser | Tech- Data and with SWF – with SWF –
36 Browser Online ***
org/lodlive/ [58] Linked Jazz | user Structure Graphics interactive
LodLive
 ACCEPTED MANUSCRIPT

Information
Output Application Access
Site URL Title of the site User need of the Input mode Comment
type category mode
user

Transcript
https://linkedjazz.
Analyser | Lay- Data and no SWF – no SWF –
37 org/network/?mo Browser Online ***
Linked Jazz | user Structure Graphics interactive
de=dynamic 59]
network

PT
*(Domain-specific). A Geographical database which delivers place names, postal codes, map views, wikipedia
info etc.

**(Domain-specific). Public opinion repository, for "Search-Rate-Compare-Choose" something from some

RI
category.

*** (Domain-Specific) A tool, part of "The Linked Jazz Transcript Analyzer" toolset project.

SC
(https://linkedjazz.org). An analyzer that identifies named entities in digital archival documents. The analysis is
applied to interview transcripts of jazz music artist and creates their network of relations and acquaintances.

4. Process of the dataset and presentation

U
For the further process of the dataset, we implemented a special application [8]. With it, is facilitated the
AN
grouping and the presentation of the dataset elements, according to combinations of any values of its dimensions
(attributes). The combinations are entered manually by the user, with the use of check-box arrays [9]. The
dataset is queried against these combinations and the matching elements of the dataset are presented in a
M

comprehensive table similar to Table 1, yet more elegant. Grouping of the dataset elements is used to create
comparative tables, where the elements of the dataset are presented according to the special characteristics,
supported by the user interfaces. In practice, the criteria of the grouping are translated into the actual filters,
D

through which the grouping is implemented. Moreover, at [9] , the reader may create a personalized
presentation, where he can group applications according to his own criteria.
TE

This special application [9] is actually a Semantic Web application. The dataset was semantically annotated in
compliance to the formalism of the model introduced in section 3. After the semantic annotation process, the
dataset was exported as an RDF file [60] . The application queries this RDF file through a SPARQL endpoint.
EP

The semantic annotation of the dataset, was achieved with the use of Zotero [61]. Zotero identifies it self as a
free tool that helps collect, organize, cite, and share research sources. The individual findings of the field
research, were maintained as Zotero collection. At first, was used the Zotero feature of the “custom tags” for the
semantic annotation of the individual reviewed web applications. The values used in the “custom tags”, were
C

selected out of the model’s attributes possible values, which are described in section 3.2. After the semantic
AC

annotation process, the whole semantically annotated collection was exported to an RDF file [60] of the
ZoteroRDF vocabulary .

4.1 Overview of the dataset

A brief quantitative description of the dataset is presented with the help of Tables 2, 3 and 4. These tables may
provide an overview and a better feeling of the dataset presented in Table 1.

Table 2.The distribution of the application categories, across the users

Application category
 ACCEPTED MANUSCRIPT

Browser Search SPARQL Tool

engine query
builder

4 13 1 2
Lay-user
applications applications application applications

10 2 1 4
Tech-user

PT
applications applications application applications

RI
If the search engines for both the Lay and Tech user (13 +2=15), are kept aside , (since they are thought to be the
standard information seeking tool), then remain 22 other non “search engine” applications. We can narrow the
scope further just to the applications whose “Output type” is not of the Semantic Web's formalism kind. This

SC
scope narrowing may provide a rough indication of how the Lay-user is served and the appropriate filter for this
is described in the Table 3.

Table 3. The plot of a filter, for non “search engine” applications and with an output free of the Semantic Web

U
formalism
AN
Attribute (Dimension) Value

“Application category” “Browser” or “SPARQL

query builder” or “Tool”
M

AND
D

Attribute (Dimension) Value

TE

“Output type” “Documents” or “no SWF –

Text” or “no SWF –
Graphics” or “no SWF –
Text & Graphics”
EP

After applying the filter described in the Table3, there are returned only seven applications which are
C

summarized in Table 4.

Table4. Summary of the results which are returned, after applying the filter in Table 3
AC

Browser SPARQL query builder Tool

4 applications 1 application 2 applications

https://www.freebase.com/ [43], http://www.irisa.fr/LIS/ferre/sparklis/ospar http://www.geonames.org [56] ,
http://www.visualdataweb.org/relfinder/ klis.html [37] http://www.swotti.com/ [57]
relfinder.php [48],
http://www.dotac.info/explorer/ [46],
https://linkedjazz.org/network/?mode=d
ynamic [59]
 ACCEPTED MANUSCRIPT

In Table 3 and apart the short quantitative presentation, it was also demonstrated the reasoning of a filter. A
similar reasoning will be used for the rest of the dataset processing, presented in the following subsections.

4.2 Criteria and formulation of the filters

The principal of the grouping is based on the model's concept. Compared to the access of just the World Wide
Web, this principal focuses on bringing forward the differentiation which should be accounted additionally,
when the access of the Semantic Web is considered. The objectives of the grouping are:

PT
• To easily distinguish the user interfaces which help in the interpretation of the data relations (in a way
closer to the human language and closer to the human way of reasoning).

RI
• To easily expose the user interfaces who help to access the data structure and their relation.

The special characteristics that a user interface needs to support will form the criteria on which the grouping of

SC
the dataset elements will be based. These criteria will further be translated into the actual filters, through which
the grouping will be implemented in practice.

The criteria will be:

U
1. User interfaces whose output (data/structures) is free of the Semantic Web formalism.
AN
2. User interfaces which return only documents. The documents are considered in a wide sense: html,
multimedia files, pdf… etc. By design, these documents do not carry the Semantic Web formalism.

3. User interfaces whose output is delivered with the Semantic Web formalism and are used for browsing
M

or navigating or exploring Semantic Web data.

4. User interfaces whose output is free of the Semantic Web formalism and are used for browsing or
navigating or exploring Semantic Web data.
D

The above criteria, are respectively translated into filters, as following :

TE

Filter 1 (for the criterion 1)

Attribute (Dimension):“Output type”
Value: [(“no SWF – Text”) OR (“no SWF – Graphics”) OR (“no SWF – Text & Graphics”)]
EP

Filter 2 (for the criterion 2)

Attribute (Dimension):“Output type”
Value:“Documents”
C

Filter 3 (for the criterion 3)

Attribute (Dimension):“Application category”
AC

Value:“Browser”
AND
Attribute (Dimension):“Output type”
Value: [(“with SWF – Graphics”) OR (“with SWF – Text & Graphics”)]

Filter 4 (for the criterion 4)

Attribute (Dimension):“Application category”
Value:“Browser”
AND
Attribute (Dimension):“Output type”
Value: [(“no SWF – Graphics” ) OR (“no SWF – Text & Graphics”)]
 ACCEPTED MANUSCRIPT

Filter 1 and filter 2 focus on the interpretation of the data relations. Filter 3 and filter 4 focus on the access of the
data structure and their relation.

4.3 Application of the filters and presentation

The application of filters 1,2,3 and 4 returns the respective groups (subsets) which are summarized in table 5.
The left column of table 5 is the filter applied, while the right column is the group (subset). It is reminded that
table 1 summarizes the individual elements of the dataset.

PT
Table 5. Summary of the application of the filters

Filter Group (subset)

RI
Filter 1 http://eol.org/ [25] , http://www.dotac.info/explorer/ [46] , http://www.geonames.org/ [56],
http://www.irisa.fr/LIS/ferre/sparklis/osparklis.html [37],

SC
http://www.visualdataweb.org/relfinder/relfinder.php [48],
https://linkedjazz.org/network/?mode=dynamic [59], https://www.freebase.com/ [43],
https://www.wikidata.org/wiki/Wikidata:Main_Page [45]

U
Filter 2 http://discoveryhub.co/ [55] , http://sisinflab.poliba.it/led/ [53] , http://www.bing.com/ [32] ,
http://www.cluuz.com/ [33] , http://www.exalead.com/search/ [35] , http://www.factbites.com/
AN
[36] , http://www.kngine.com/ [38] , http://www.sensebot.net/ [40], http://www.swotti.com/
[57] , https://duckduckgo.com/ [41] , https://ixquick.com/ [42] , https://www.startpage.com/
[44]
M

Filter 3 http://code.know-center.tugraz.at/search [54] , http://en.lodlive.it/ [24] , http://lodview.it/ [27]

, http://ws.nju.edu.cn/falcons/objectsearch/index.jsp [31] ,
D

http://www.visualdataweb.org/gfacet/gfacet.php [47] ,
http://www.visualdataweb.org/semlens/semlens.php [49] ,
TE

http://www.visualdataweb.org/tfacet/tfacet.php [50] , https://linkedjazz.org/lodlive/ [58]

Filter 4 http://www.dotac.info/explorer/ [46] , http://www.visualdataweb.org/relfinder/relfinder.php

[48] , https://linkedjazz.org/network/?mode=dynamic [59] , https://www.freebase.com/ [43]
EP

After the application of filters 1,2,3 and 4, we note the following observations:
C

For the results returned after the application of filter 1: (User interfaces whose output is free of the Semantic
Web formalism). As expected they all are suitable for the Lay-user. Their categories are of all four kinds (Search
AC

engine, Browser, Tool, even a SPARQL query builder is included). The information needs served, are data or
data and structured combined. The input mode, if not interactive, is just a search term form. In any case the
input mode is also free of the Semantic Web formalism. All these interfaces are accessible only online.

For the results returned after the application of filter 2: (It concerns interfaces which return only documents
such as html, multimedia files, pdf… etc.). The documents themselves are not of the Semantic Web discipline
and therefore the information need served by the documents is the one for data. These interfaces are suitable for
the Lay-user. These interfaces are search engines except [57] , which is considered as “tool” since the
documents it returns are not discovered by crawling the SW (similarly to the crawling of the WWW) but are
contributed by the users. All though these interfaces return documents, there is one, the "Cluuz Search" [33] ,
which apart the documents, returns a collection of graphs with related entities (relations discovered through the
processing of the semantically annotated linked data). Another notable interface in this group, is the “Discovery
 ACCEPTED MANUSCRIPT

Hub | Beta” [55] which provides an interesting alternative “query suggestion” feature. All these interfaces are
accessible only online.

For the results returned after the application of filter 3: (Interfaces whose output is delivered with the
Semantic Web formalism and are used for browsing or navigating or exploring). As expected these are suitable
for the Tech-users. The information needs served, are data or structure or data and structured combined. The
input mode is also of the Semantic Web discipline, the same as the output type. An exception in the input type, is
the “CODE Linked Data Query Wizard” [54] which uses a search term form.

PT
For the results returned after the application of filter 4: (Interfaces whose output is free of the Semantic Web
formalism and are used for browsing or navigating or exploring). This group is a collection from the opposite
side of filter 3. It should be noted that this group is relatively short (only four elements) which implies a limited

RI
availability of Semantic Web browsers suitable for Lay-users.

5. Discussion and conclusions

SC
In this work we conducted a review of applications which provide a user interface for accessing Semantic Web
data and their relations. The findings of the field research were assembled into a set, according to a six-

U
dimensional model (see Section 3). The elements of this set, (the individual applications) were studied and
presented comparatively and were evaluated against those special characteristics that a user interface needs to
AN
support, with respect to its application in the access of the Semantic Web.

Since the user interface is in focus, the user is the centre of the application's design considerations. It is worth
noting that the first dimension of the model, is the “User” and this dimension may be assigned the value “Tech-
M

user“(users who understand the Semantic Web and are familiar with the related technologies) or the value “Lay-
user” (users who know how to use a computer, without necessarily understanding the Semantic Web). In the
context of the Semantic Web, it is the user who has to interpret the meaning (the semantics) of the returned data,
D

if it is not the machine itself.

In the introduction of the Semantic Web vision [1] and in two of its follow-ups [2, 3] the agents will be bearing
TE

the functionality of the semantics interpretation and the interaction with the user. However, since the agents are
not yet deployed in large scale, these needs are still served by the user interfaces, such like those studied in this
review.
EP

At present, most of the Semantic Web applications are information seeking. This work is based on the
assumption that when a user of an application interacts with the Semantic Web, through any kind of a user
interface, this is done mostly to satisfy information needs and less for processing existing data or for adding
C

new. Further to this, the general impression is that the support of the Lay-user is limited, compared to this of the
Tech-user.
AC

Some of the reviewed applications/user interfaces are effective in the data/relations visualization and in the
interaction with the user (i.e. [24, 47, 50, 51] ) but they are targeting the Tech-user. Of course, those targeting the
Lay-user are also very interesting on all aspects (i.e. [37, 48]).

A special segment is this of the search engines. The concept of the search engines is focusing on delivering
documents (html, multimedia files, pdf… etc.) regardless the data collections and the data structure/relations.
However in the search engine in [33] , and since the Semantic Web technology can span more aspects than just
returning ranked documents, the user interface visualizes its findings through niche graphs, relevant to the
search term or search phrase.

The Semantic Web technology with its inherent capabilities of data linking and data relations representation,
may be used to infer what is relevant to a certain information need and also what is not relevant. This dual
functionality is the core concept of the [55] which suggests other data that might be of interest, yet not very
 ACCEPTED MANUSCRIPT

relevant. Furthermore, this triggers some concern about the way relevance should be judged in the Semantic
Web, compared to the WWW.

All though the upper layers of the Semantic Web stack are not yet fully implemented, it is still possible to
deliver applications for the Lay-user as already found. The abstraction of the Semantic Web formalism, together
with the interaction in a language close to the natural one, are the key points for developing Semantic Web
applications for the Lay-user. The “Linked Jazz Transcript Analyser" (https://linkedjazz.org) stands as a
characteristic example, as if its two variants resemble the facets of the same coin. The “Linked Jazz Transcript

PT
Analyser" is a project of a very specific domain. It is an analyser that identifies named entities in digital archival
documents. These documents are interview transcripts of jazz music artists. The analysis of these transcripts
creates and visualizes the network of relations and acquaintances of the jazz music artists. The project delivers

RI
two applications which perform the same task but with different appearance (with different user interface).
These are the “Transcript Analyzer | Linked Jazz | LodLive” [58] (https://linkedjazz.org/lodlive/) and the
“Transcript Analyzer | Linked Jazz | network” [59] (https://linkedjazz.org/network/?mode=dynamic ). The

SC
“Lodlive” version is the one for the Tech-user while the “network” version is the one for the Lay-user. If these
two versions are considered comparatively, they are a very interesting showcase on the difference of the user
experience, as far as the interpretation of the data and their relation is concerned. In the “Lodlive” version which
requires the formalism of the Semantic Web, the entities to query are entered as URIs and the network is

U
returned with the same formalism. In the “network” version, the entities to query are entered as keywords
selected out of a list and the network is returned in a form that any user can comprehend easily. In any case there
AN
is also the option to bring up relevant documents and text information with a mouse click. Screenshots showing
the network of the exact same entity (the jazz musician Nat King Cole) and also demonstrating the difference of
the user experience between the “Transcript Analyzer | Linked Jazz | LodLive” and the “Transcript Analyzer |
M

Linked Jazz | network”, are presented in Figure 2 and Figure 3 respectively

D
TE

Figure 2 (place holder)

File name “Figure2_Transcript Analyzer_lodlive.tiff”
Figure is embedded in the document in its original size,
and then is scaled by the application
EP

(zoom-in brings up the details)

C

Figure 2. Nat King Cole's network, delivered through the “Transcript Analyzer | Linked Jazz | LodLive” tool
AC

Figure 3 (place holder)

File name “Figure3_Transcript Analyzer_network.tiff”
Figure is embedded in the document in its original size,
and then is scaled by the application
(zoom-in brings up the details)
 ACCEPTED MANUSCRIPT

Figure 3. Nat King Cole's network, delivered through the “Transcript Analyzer | Linked Jazz | network” tool

6. Future work

PT
Having in mind the findings of this review, some further contribution might be:

The improvement and extension of the model introduced in this work. This could lead to the creation of a

RI
reusable RDF vocabulary or an OWL ontology for use in further study of the Semantic Web user interfaces.
Other potential uses of such an RDF vocabulary or an OWL ontology might be the i) attributing and cataloging
of Semantic Web user interfaces, and ii) distribution of the metadata required by applications, for the

SC
optimization of their user interface either at run or at design time.

To continue the search for adding new elements in the collection of Semantic Web applications. The application
built by the authors in order to maintain this collection, might even evolve to a community curated dataset.

U
The training of the Lay-user. The standardization of methodologies for the navigation, exploration, information
AN
searching and SPARQL query building, for the purpose of familiarizing the Lay-user with the Semantic Web.

An approach to tapping the Semantic Web, with the building of user interfaces (i.e. software agents) to which the
user will be interacting in a way closer to the way the humans communicate between them. The basic
M

components for such an approach are already available. Frameworks such the “JADE” [62] for the
implementation of multi-agent systems, the “Apache Jena” [63] which delivers RDF/OWL API, SPARQL
endpoint, triple store and inference engine or the “OWL API” [64] for accessing OWL ontologies, can be used
D

to produce applications who are Semantic Web enabled. These frameworks used together with NLP frameworks
might lead to applications who could add-up a little bit to the current top of the Semantic Web stack.
TE

Security considerations for the “SPARQL 1.1 Update” [65] .The Semantic Web is not “read-only”. The
“SPARQL 1.1 Update” is a language for specifying and executing updates to RDF files. Eventually, over the
time, this functionality will become a required feature of the user interface. Since RDF files will be “read/write”
EP

at the will of the user, there will emerge several security issues [66] (i.e. SPARQL injection) and also the need
for counter measures incorporated in the user interface.
C

References
AC

[1] T. Berners-Lee, J. Hendler, and O. Lassila, “The Semantic Web,” Scientific American, pp. 29–37, May-
2001.
[2] N. Shadbolt, W. Hall, and T. Berners-Lee, “The Semantic Web Revisited,” Intell. Syst. IEEE, vol. 21, no. 3,
pp. 96–101, Jan. 2006.
[3] C. Bizer, T. Heath, and T. Berners-Lee, “Linked data - The story so far,” Int. J. Semantic Web Inf. Syst., vol.
5, no. 3, pp. 1–22, 2009.
[4] w3c, “layerCake-4.png (PNG Image),” 10-May-2016. [Online]. Available:
https://www.w3.org/2007/Talks/0130-sb-W3CTechSemWeb/layerCake-4.png. [Accessed: 10-May-2016].
[5] “Semantic Web Case Studies and Use Cases,” Semantic Web Case Studies and Use Cases, 10-May-2016.
[Online]. Available: https://www.w3.org/2001/sw/sweo/public/UseCases/. [Accessed: 10-May-2016].
[6] I. Dickinson, “Position statement: in favour of (more) intelligence in the semantic UI,” in SWUI 2009
Sharing Ideas for Complex Problems in User Interaction, Washington DC, USA, 2009, vol. Vol-544.
 ACCEPTED MANUSCRIPT

[7] A. Nanavati, “User Interfaces for Semantic Web Applications,” The University of Texas at Austin, 03-May-
2005. [Online]. Available: https://www.ischool.utexas.edu/~i385q/archive/nanavati_a/semweb_UI.htm.
[Accessed: 10-May-2016].
[8] Charalampos C. Charalampidis and Euclid A. Keramopoulos, “Modeling the semantic web user interface -
Case study on building a semantic web application,” Manuscr. Submitt. Publ., Sep. 2016.
[9] Charalampos C. Charalampidis and Euclid A. Keramopoulos, “Semantic Web User Interfaces,” 2016.
[Online]. Available: http://sw.it.teithe.gr/apps/swui/. [Accessed: 12-Jun-2016].
[10] M. Martin and S. Auer, “Categorisation of Semantic Web Applications,” in SEMAPRO 2010, The Fourth

PT
International Conference on Advances in Semantic Processing, Florence, Italy, 2010, p. 179 to 185.
[11] “W3C Semantic Web FAQ,” 10-May-2016. [Online]. Available: https://www.w3.org/RDF/FAQ. [Accessed:
10-May-2016].
[12] A.-S. Dadzie and M. Rowe, “Approaches to visualising Linked Data: A survey,” IOS Press, Jan. 2011.

RI
[13] A. Khalili and S. Auer, “WYSIWYM – Integrated visualization, exploration and authoring of semantically
enriched un-structured content,” Semantic Web, vol. 6, no. 3, pp. 259–275, Jan. 2015.
[14] S. Mazumdar, D. Petrelli, K. Elbedweihy, V. Lanfranchi, and F. Ciravegna, “Affective graphs: The visual

SC
appeal of Linked Data,” Semantic Web, vol. 6, no. 3, pp. 277–312, Jan. 2015.
[15] A. Azizan, Z. A. Bakar, N. K. Ismail, and M. F. Amran, “Interface features of semantic web search engine,”
in e-Learning, e-Management and e-Services (IC3e), 2013 IEEE Conference on, 2013, pp. 142–147.
[16] C. D. Manning, Ρ. Raghaνan, and H. Schütze, Introduction to information retrieval. Cambridge University
Press, 2008.

U
[17] G. Antoniou and F. van Harmelen, A Semantic Web Primer, Second edition. The MIT Press, 2008.
[18] P. Frischmuth, M. Martin, S. Tramp, T. Riechert, and S. Auer, “OntoWiki – An authoring, publication and
AN
visualization interface for the Data Web,” Semantic Web, vol. 6, no. 3, pp. 215–240, Jan. 2015.
[19] E. Turner, A. Hinze, and S. Jones, “A Review of User Interface Adaption in Current Semantic Web
Browsers.” The University of Waikato, 08-Feb-2011.
[20] P. Hoefler, M. Granitzer, E. Veas, and C. Seifert, “Linked data query wizard: A novel interface for accessing
M

sparql endpoints,” presented at the CEUR Workshop Proceedings, 2014, vol. 1184.
[21] J. Nielsen, “Usability 101: Introduction to Usability,” Usability 101: Introduction to Usability, 04-Jan-
2012. [Online]. Available: https://www.nngroup.com/articles/usability-101-introduction-to-usability/.
D

[Accessed: 10-May-2016].
[22] N. Marie and F. Gandon, “Survey of linked data based exploration systems,” in IESD 2014 - Intelligent
Exploitation of Semantic Data, Riva Del Garda, Italy, 2014.
TE

[23] “Welcome - the Datahub,” 10-May-2016. [Online]. Available: https://datahub.io/. [Accessed: 10-May-
2016].
[24] “LodLive - browsing the Web of Data,” 10-May-2016. [Online]. Available: http://en.lodlive.it/. [Accessed:
10-May-2016].
EP

[25] “Encyclopedia of Life - Animals - Plants - Pictures & Information,” 10-May-2016. [Online]. Available:
http://eol.org/. [Accessed: 10-May-2016].
[26] “Q&D RDF Browser,” 10-May-2016. [Online]. Available: http://graphite.ecs.soton.ac.uk/browser/.
[Accessed: 10-May-2016].
C

[27] “LodView — giving data a new shape,” 10-May-2016. [Online]. Available: http://lodview.it/. [Accessed:
10-May-2016].
AC

[28] “OpenLink Data Explorer Extension,” 20-May-2016. [Online]. Available: http://ode.openlinksw.com/.

[Accessed: 20-May-2016].
[29] “Swoogle Semantic Web Search Engine,” 10-May-2016. [Online]. Available: http://swoogle.umbc.edu/.
[Accessed: 10-May-2016].
[30] “Virtuoso SPARQL Query Editor,” 20-May-2016. [Online]. Available: http://virit.it.teithe.gr/sparql/.
[Accessed: 20-May-2016].
[31] “Falcons Object Search,” 10-May-2016. [Online]. Available:
http://ws.nju.edu.cn/falcons/objectsearch/index.jsp. [Accessed: 10-May-2016].
[32] “Bing,” 10-May-2016. [Online]. Available: http://www.bing.com/#. [Accessed: 10-May-2016].
[33] “Cluuz Search,” 10-May-2016. [Online]. Available: http://www.cluuz.com/. [Accessed: 10-May-2016].
[34] N. Humfrey, “EasyRdf - RDF Library for PHP,” EasyRdf - RDF Library for PHP, 10-May-2016. [Online].
Available: http://www.easyrdf.org/. [Accessed: 10-May-2016].
[35] “Web Search - Exalead,” 10-May-2016. [Online]. Available: http://www.exalead.com/search/. [Accessed:
10-May-2016].
 ACCEPTED MANUSCRIPT

[36] “Factbites: Where results make sense,” 10-May-2016. [Online]. Available: http://www.factbites.com/.
[Accessed: 10-May-2016].
[37] “Sparklis,” 10-May-2016. [Online]. Available: http://www.irisa.fr/LIS/ferre/sparklis/osparklis.html.
[Accessed: 10-May-2016].
[38] “Kngine - Ask me anyhing,” 10-May-2016. [Online]. Available: http://www.kngine.com/. [Accessed: 10-
May-2016].
[39] “Twinkle: A SPARQL Query Tool,” 10-May-2016. [Online]. Available:
http://www.ldodds.com/projects/twinkle/. [Accessed: 10-May-2016].

PT
[40] “SenseBot - semantic search engine that finds sense on the Web,” 10-May-2016. [Online]. Available:
http://www.sensebot.net/. [Accessed: 10-May-2016].
[41] “DuckDuckGo,” 10-May-2016. [Online]. Available: https://duckduckgo.com/. [Accessed: 10-May-2016].
[42] “Ixquick Search Engine,” 10-May-2016. [Online]. Available: https://ixquick.com/. [Accessed: 10-May-

RI
2016].
[43] “Freebase,” 10-May-2016. [Online]. Available: https://www.freebase.com/. [Accessed: 10-May-2016].
[44] “Startpage Web Search,” 10-May-2016. [Online]. Available: https://www.startpage.com/. [Accessed: 10-

SC
May-2016].
[45] “Wikidata,” 10-May-2016. [Online]. Available: https://www.wikidata.org/wiki/Wikidata:Main_Page.
[Accessed: 10-May-2016].
[46] “Tim Berners-Lee :: dotAC Explorer,” 10-May-2016. [Online]. Available: http://www.dotac.info/explorer/.
[Accessed: 10-May-2016].

U
[47] “gFacet,” 10-May-2016. [Online]. Available: http://www.visualdataweb.org/gfacet/gfacet.php. [Accessed:
10-May-2016].
AN
[48] “RelFinder,” 10-May-2016. [Online]. Available: http://www.visualdataweb.org/relfinder/relfinder.php.
[Accessed: 10-May-2016].
[49] “SemLens,” 06-Jan-2016. [Online]. Available: http://www.visualdataweb.org/semlens/semlens.php.
[Accessed: 06-Jan-2016].
M

[50] “tFacet,” 06-Jan-2016. [Online]. Available: http://www.visualdataweb.org/tfacet/tfacet.php. [Accessed: 06-

Jan-2016].
[51] “OpenLink iSPARQL,” 20-May-2016. [Online]. Available: http://virit.it.teithe.gr/isparql/. [Accessed: 20-
D

May-2016].
[52] “SView Homepage,” 10-May-2016. [Online]. Available: http://ws.nju.edu.cn/sview/. [Accessed: 10-May-
2016].
TE

[53]: : “LED :: Lookup Explore Discover,” 10-May-2016. [Online]. Available: http://sisinflab.poliba.it/led/.

[Accessed: 10-May-2016].
[54] “CODE Linked Data Query Wizard,” 10-May-2016. [Online]. Available: http://code.know-
center.tugraz.at/search. [Accessed: 10-May-2016].
EP

[55] “Discovery Hub | Beta,” 10-May-2016. [Online]. Available: http://discoveryhub.co/. [Accessed: 10-May-
2016].
[56] “GeoNames,” 10-May-2016. [Online]. Available: http://www.geonames.org/. [Accessed: 10-May-2016].
[57] “Swotti Opinion Analizer. Search, Rate and Compare,” 10-May-2016. [Online]. Available:
C

http://www.swotti.com/. [Accessed: 10-May-2016].

[58] “Transcript Analyzer | Linked Jazz | LodLive,” 10-May-2016. [Online]. Available:
AC

https://linkedjazz.org/lodlive/. [Accessed: 10-May-2016].

[59] “Transcript Analyzer | Linked Jazz | network,” 10-May-2016. [Online]. Available:
https://linkedjazz.org/network/?mode=dynamic. [Accessed: 10-May-2016].
[60] “RDF data-set (Semantic Web user interfaces),” 12-Jun-2016. [Online]. Available:
http://sw.it.teithe.gr/data_rdf/swui.rdf. [Accessed: 12-Jun-2016].
[61] “Zotero,” 10-May-2016. [Online]. Available: http://zotero.org/. [Accessed: 10-May-2016].
[62] “Jade Site | Java Agent DEvelopment Framework,” 10-May-2016. [Online]. Available:
http://jade.tilab.com/. [Accessed: 10-May-2016].
[63] “Apache Jena - Home,” 10-May-2016. [Online]. Available: https://jena.apache.org/. [Accessed: 10-May-
2016].
[64] “OWL API,” 10-May-2016. [Online]. Available: http://owlapi.sourceforge.net/. [Accessed: 10-May-2016].
[65] “SPARQL 1.1 Update,” 10-May-2016. [Online]. Available: http://www.w3.org/TR/sparql11-update/.
[Accessed: 10-May-2016].
 ACCEPTED MANUSCRIPT

[66] “SPARQL 1.1 Update /#security,” 10-May-2016. [Online]. Available: https://www.w3.org/TR/sparql11-

update/#security. [Accessed: 10-May-2016].

PT
RI
U SC
AN
M
D
TE
C EP
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M
D
TE
EP
C
AC