SEARCH ENGINE IN A CLASS OF ACADEMIC

DIGITAL LIBRARIES

Maria Nisheva-Pavlova, Pavel Pavlov

Faculty of Mathematics and Informatics, Sofia University

5 James Bourchier Blvd., Sofia 1164, Bulgaria
e-mail: {marian, pavlovp}@fmi.uni-sofia.bg

Abstract
The paper discusses some aspects of an ongoing project aimed at the
development of a methodology and proper software tools for building and
usage of academic digital libraries. A particular functional model of academic
digital library has been proposed and analyzed. The emphasis falls on some
solutions of the large set of problems concerning the development of
adequate mechanisms for semantics oriented search in multilingual digital
libraries. An ontology-based approach is suggested in order to standardize
the semantic annotation of the library resources and to facilitate the
implementation of the functionality of the search engine. The main features of
a prototype of knowledge-based search engine for a multilingual academic
digital library with research and learning materials are discussed. This search
engine uses proper ontologies describing the conceptual knowledge
considerable for the chosen domains and in this way it is capable of retrieving
and filtering documents by their semantic properties.

Keywords: Digital Library; Metadata; Semantic Annotation; Ontology;

Search Engine

1. Introduction

Research and practical activities in the field of Digital Libraries during the last
two decades lead to significant results in the development and management
of digital collections, in the innovation in scholarly publishing and the long-
term preservation of digital information. Many institutions are actively
involved in building suitable repositories of the institution's books, papers,
theses, and other works which can be digitized or were “born digital”. In

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

particular, universities and other academic institutions participate

successfully in lots of projects directed to the development of different types
of academic digital libraries. Academic digital libraries are committed to
maintaining valuable collections of scholarly information. To this end,
essential information resources should remain available and accessible into
the future – a real challenge in the cases of digital resources that are
increasingly transient and at risk.
The paper is aimed at the presentation of an ongoing project which is
directed to the development of a methodology and corresponding software
tools for building academic digital libraries. A special attention has been paid
to the elaboration of means for semantics oriented search in multilingual
digital libraries. The study and the practical experiments are oriented to the
development of DigLib-CI – a digital library with research and learning
materials (articles, dissertations, monographs, lecture notes, textbooks,
presentations, example program sources, data sets, quizzes, manuals etc.)
created at the Department of Computer Informatics of the Faculty of
Mathematics and Informatics (FMI), Sofia University, or especially selected
among the scholarly materials freely available on the Web.

2. Related Work

Digital Libraries can mainly be characterized as a converging point where

disparate communities have been meeting to address common issues related
with the creation, management and usage of digital information [1]. The goal
of a digital library and especially of an academic library is to provide access to
selected intellectual works. Moreover, academic digital libraries are usually
aimed at some specific challenges like digital preservation of valuable
scientific heritage collections and investigation of innovative methods for
automatic indexing, metadata extraction, document search and retrieval etc.
In this sense, academic digital libraries are the front-rankers in the discussed
area.
The digital libraries of Cornell University [2], the University of Michigan
[3] and Carnegie Mellon University [4] are considered as leaders in the field of
academic digital library creation and management.
The Cornell University Library is the eleventh largest academic library in
the United States, ranked by number of volumes held. In 2005 it held 7.5
million printed volumes in open stacks, 8.2 million microfilms and
microfiches, and a total of 440,000 maps, motion pictures, DVDs, sound

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Paper Title

recordings, and computer files in its collections, in addition to extensive

digital resources and the University Archives.
The Cornell Library Digital Collections Project integrates online collections
of historical documents. Featured collections include the Database of African-
American Poetry, the Historic Math Book Collection, the Samuel May Anti-
Slavery Collection, the Witchcraft Collection, and the Donovan Nuremberg
Trials Collection.
The University of Michigan Digital Library Project (UMDL) is based on
the traditional values of service, organization, and access that have made
libraries powerful intellectual institutions in combination with open,
evolving, decentralized advantages of the web. The content of UMDL will
emphasize a diverse collection, focused on earth and space sciences, which
can satisfy the needs of many different types of users. The content will be
supplied by publishers, although the project will eventually allow all users to
publish their work.
The implementation of the current prototype of UMDL requires the
integration of numerous agent technologies for knowledge exchange,
commerce, learning, and modelling. Recently, the efforts have been
concentrated on developing technologies that, for example, manipulate
ontological descriptions of the elements of a digital library to help agents to
find services and auctions for exchanging goods and services under various
conditions. These technologies allow flexibility in the UMDL configuration
policies, extensibility and scalability by using demand as incentive for
replicating services.
Carnegie Mellon University Libraries became very popular with the
Million Book (or the Universal Library) project which was aimed to digitize a
million books by 2007. The activities within the project include scanning
books in many languages, using OCR to enable full text searching, and
providing free-to-read access to the books on the web. As of 2007, they have
completed the scanning of the planned number of books and have made
accessible the corresponding database.
The research within the Million Book project includes developments in
machine translation, automatic summarization, image processing, large-scale
database management, user interface design, and strategies for acquiring
copyright permission at an affordable cost.
Compared to these well-known large scale initiatives, our project is of a
significantly smaller scale, but in contrast to all of them, it investigates the use
of a set of subject ontologies to provide flexible, semantics-oriented access to
the library resources for users with different profiles and language skills.

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

3. Architecture of DigLib-CI

DigLib-CI is designed as a typical academic digital library. It has been under

development at FMI in order to provide open access to various kinds of
scholarly and instructional content, mainly in a wide range of subfields of
Computer Science and Information Systems. The functional structure of
DigLib-CI is shown in Figure 1.
The content repositories include research and learning materials of
different types (books, dissertations, periodicals and single articles, manuals,
lecture notes, presentations, source code of computer programs, data sets,
tests, quizzes etc.) in the areas of Computer Science and Information Systems.
These library resources are available in various digital formats: pdf, html,
plain text, doc, ppt, jpeg etc. Most of them are developed by faculty members,
the others are especially selected among the scholarly materials freely
available on the Web. The content repositories are stored in a small number of
locations. The materials in them are written in Bulgarian or in English
language.
The metadata catalogues are destined to facilitate the identification of the
needed research or learning materials by the search engine. They contain
descriptive metadata stored in XML format and support the reusability of all
library resources and facilitate their interoperability.

GUI
Library Functionality

Search Engine
Subject Ontologies and Index

Metadata Metadata Metadata

Catalogue Catalogue Catalogue

Content Content ... Content

Repository Repository Repository

Figure 1: Functional Model of DigLib-CI

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Paper Title

The subject ontologies include large sets of concepts of the areas of

Computer Science and Information Systems, with description of their
properties and different kinds of relationships among them. They play a
significant role in the implementation of the full functionality of the search
engine.
The purpose of the search engine is to provide adequate access to the
complete palette of resources stored in DigLib-CI.
The library functionality and the user interface of DigLib-CI are designed
in accordance with the expected needs and requirements of the basic types of
users of the library. The interface module provides adequate online access to
the corresponding library resources and supporting software tools.

4. Catalogue Metadata

The library catalogues contain metadata which support the identification of

the requested resources by the search engine. These metadata are stored in
XML format and comply with the IEEE Standard for Learning Object
Metadata [5].
Typical examples of relevant attributes of most kinds of research and
learning materials are: type of the material; author; title of the material;
language(s) (human and/or programming one(s)); digital format; location;
version; date of creation; completion status; restrictions on use; semantic
annotation – list of concepts from a proper subject ontology describing the
Computer Science or Information Systems subfields and/or concepts covered
or treated by the material. Learning materials have been characterized also by
their educational level and the principal types of users for which the
corresponding material was designed; officially published research materials
and textbooks are supplied with the corresponding bibliographic metadata.
Each catalogue entry (i.e., each resource description) consists of two
equivalent parts in which the element values are texts in Bulgarian or English
language respectively. The search engine examines the corresponding parts of
the descriptions according to the language of the user query.
The elements <ontologyRefs> and <keywords> of the resource descriptions
play the role of semantic annotations of the corresponding library materials.
The values of the child elements of <ontologyRefs> are concepts of the
suitable subject ontologies (names of classes in these subject ontologies) which
present most precisely the content of the corresponding document.

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

The concepts of the subject ontologies are too general from the point of
view of the expectations of the typical users of DigLib-CI. For that reason one
can include in the resource descriptions additional lists of keywords which
describe the content of the corresponding documents at the necessary level of
abstraction. These keywords are set as values of the child elements of the
<keywords> resource description elements.
The names of the respective subject areas and names of the files containing
the suitable subject ontologies have been assigned as values of the child
elements of the catalogue description elements <subjects> and <ontologies>
respectively.

5. Subject Ontologies

The subject ontologies include a large set of concepts in the fields of

Computer Science and Information Systems, with description of their
properties and the different kinds of relationships among them. Two subject
ontologies are included in the current version of DigLib-CI. The Computer
Science ontology is based on the Computer Science Curriculum 2008 of ACM
and IEEE/CS [6]. Using the curriculum as a guideline, this ontology defines
the atomic knowledge units for the University courses and available research
materials in the field of Computer Science and makes them sharable and
reusable. Its current version includes approximately 300 concepts with their
relationships.
The Information Systems ontology has been under development using the
Model Curriculum and Guidelines for Undergraduate Degree Programs in
Information Systems of ACM, AIS and AITP [7].
The subject ontologies are designed in order to play the role of
information sources describing the hierarchy and the other relationships
between the main concepts in the discussed domains. A dictionary of
synonyms has also been under development with the purpose of providing
the search engine with other viewpoints to the conceptual structure of the
areas of Computer Science and Information Systems.
The body of knowledge in the areas of Computer Science and Information
Systems is formulated in the terms of a considerable number of common
concepts, therefore the two subject ontologies discussed above contain many
common classes (with equal or similar names and intersecting properties and
restrictions on them). Because of that our further plans include the
development of an approach to the integration of domain ontologies relevant

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Paper Title

to the contents of multilingual academic digital libraries which will be based

on some of our former results [8].

6. User Interface

The library functionality and the user interface of DigLib-CI are designed in
accordance with the expected requirements of the basic types of users of the
library. The interface module provides adequate online access to the
corresponding library resources and supporting software tools.
The current version of the user interface allows one to formulate queries in
Bulgarian or English language. It is intended for four types of users:
• FMI students – they may read/download textbooks, open lecture notes
and presentations from all public sections of the library as well as all
manner of other kinds of materials (monographs, dissertations, articles,
periodicals, degree theses, lecture notes, presentations, exercises,
programs, data sets, quizzes, tests etc.) from fixed public library
sections;
• FMI lecturers and researchers – in addition to the students’ access
rights, they may upload materials to fixed public sections as well as
create and update private sections and use materials in some of them;
• librarians (library administrators) – they have full access to all public
resources of the library (may download and upload materials destined
for all public sections of the library);
• general citizen – they may read and download public materials of
fixed types (e.g., dissertations, textbooks, open lecture notes and
presentations).
All types of users of DigLib-CI may use the standard input interface which
gives up convenient means for entering, editing and submitting queries for
various kinds of document search and retrieval. FMI lecturers and researchers
as well as the library administrators may play the role of authors of library
resources and have an access to the author’s part of the user interface. This
part of the user interface places at the authorized persons’ disposal
appropriate forms enabling one to enter and edit catalogue descriptions of all
types of library resources (Figure 2). More precisely, the user may enter the
values of some of the elements or pick out the values of others from
previously drawn lists. In particular, the available subject ontologies can be
properly visualized and the necessary concepts in them can be picked out as
values of the child elements of the element <ontologyRefs>.

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

Figure 2: User interface of DigLib-CI (author’s view – form for entering

catalogue metadata of periodicals)

7. Working Principles of the Search Engine

The purpose of the search engine is to provide adequate access to the

complete palette of resources stored in DigLib-CI.
The search engine maintains several types of search and document
retrieval within DigLib-CI. The user queries define restrictions on the values
of certain metadata attributes of the required research or learning materials.
Generally the search mechanism may be formulated as follows: the document
descriptions included in all permissible for the user sections of the library are
examined one by one and these descriptions which have a specific element
(determined by the type of the user query) with a value matching the user
query, are marked in order to form the search result. The matching process is
successful if the value of the element or the value of one of its child elements
is equal to the user query. The documents pointed by the marked descriptions
are retrieved and the user is given an access to these documents and their
catalogue descriptions.

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Paper Title

The current implementation of the search engine supports four types of

search and document retrieval:
• full search – search and retrieval of all available library resources,
ordered by title, by author, by category, by date of creation or by date
of inserting in the library;
• author search (search and retrieval of the documents created by a
given author) – the search is performed in the value of the element
<authors>;
• ontological search – the search is performed in the value of the element
<ontologyRefs>;
• keyword search – the search is performed in the value of the element
<keywords>.
During the ontological search the user query is augmented with regard to
the concepts searched out in the semantic annotations of the required research
or learning materials. The more specific concepts from each of the subject
ontologies indicated by the user are added to the original one in the resulting
query. Then the search engine retrieves all documents in the library
containing in their descriptions at least one component of the augmented
query as the value of a child element of <ontologyRefs>. In this way the
ontological search enables one to find documents described by ontology
concepts which are semantically related to the concept defining the user
query.
Till now, we have no disposal of an accomplished proper dictionary of
synonyms of the concepts in the areas of Computer Science and Information
Systems neither in Bulgarian, nor in English, but our idea is to provide a
possibility for two-stage augmentation of the user query. At the first stage the
request for ontological search will be extended with the more specific
concepts (its successors) from the indicated subject ontologies. At the second
stage the synonyms found in the dictionary will be added to the main (given
by the user) concept and its successors.
We allow in the current version of the implementation of the search engine
only “atomic” user queries that do not contain conjunctions or disjunctions of
words or phrases. The next step will be to elaborate a sophisticated version of
the search engine which will be capable to analyze and execute queries in the
form of conjunctions or disjunctions of phrases of interest for the user. Some
our former ideas suggested in [9] will be used for the purpose.
The discussed working principles of the search engine of DigLib-CI are
designed in order to support flexibility, interoperability and reusability. These
principles could be applied in the implementation of the search engines of a

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

whole class of academic digital libraries that provide semantics oriented

access to their resources.

8. An Example of Ontological Search

Let us suppose for example that the user defines a request (a query) for
ontological search concerning the concept “fundamental constructs”. First an
extension of this request will be generated. It will include all ontological
concepts which are special cases of the concept given by the user (with respect
to the ontologies indicated by the user). For this purpose, breadth-first search
in the graphs that represent the ontologies will be performed, starting in each
one from the concept chosen by the user.
Assume that the Computer Science ontology is chosen by the user. In this
case the extended request (the augmented query) will include the concepts
“fundamental constructs”, “basic syntax and semantics”, “binding and
scope”, “conditional structures”, “declarations”, “expressions”, “functions
and procedures”, … , “variables”, “bindings”, “blocks”, … , “simple
variables”.

Figure 3: Some search results for the query “fundamental constructs”

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Paper Title

After that, a consecutive search in the catalogue descriptions follows. In

this search all documents with descriptions that are juxtaposed with at least
one element of the extended request, are extracted. In the current
implementation each document appears as many times in the result list, as
many elements of the augmented query are juxtaposed with its description
(which means that the element <ontologyRefs> of the description includes a
sub-element that has value, coincident with an element of the augmented
query).
Figure 3 shows a screenshot displaying part of the ontological search
results for the query “fundamental constructs”.
If the user indicates more than one subject ontology (e.g., the Computer
Science ontology and the Information Systems ontology), the procedure
described above is repeated consecutively for each of these ontologies.
Our current activities are directed to the selection of a proper set of
relationships between the ontology concepts that should be taken into account
in the process of ontological search along with the hierarchical ones. We
envisage for the near future the development of a more flexible and user-
friendly mechanism for ontological search which will not expect from the user
to indicate explicitly the subject ontologies appropriate for every particular
case.

9. Conclusions

The most considerable results of the discussed project obtained so far may be
summarized as follows:
• A functional model of an academic digital library was proposed. This
model provides tools for semantics oriented access to learning and
research materials in various digital formats written in different
languages;
• A prototype of DigLib-CI – an academic digital library with research
and learning materials in the areas of Computer Science and
Information Systems, was developed.
The main advantage of the suggested approach to building academic
digital libraries consists in the provided facilities for flexible and adequate
semantics-oriented access to the library resources for users with various
professional profiles and language skills.
The complete implementation of the project will help to enhance the
research activities and the exchange of teaching innovation and thus will

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland
 Authors

improve the overall scholarly and teaching quality in Computer Science and
Information Systems at FMI. It will also contribute to the methodology of
development of innovative software systems maintaining the entire lifecycle
of academic digital content.
Acknowledgements. This work has been partly funded by the Sofia
University SRF within a project titled “Methods and Tools Supporting the
Lifecycle of Rich Digital Content”.

References

[1] BORBINHA, J. The Age of the Digital Library. In: D. Castelli, E. Fox
(Eds.), Pre-proceedings of the first International Workshop on
Foundations of Digital Libraries, Vancouver, Canada, 2007, pp. 31-36.
[2] Cornell University Library Digital Collections. Available at
http://cdl.library.cornell.edu/ (March 2010).
[3] University of Michigan Digital Library. Available at
http://www.si.umich.edu/UMDL/ (March 2010).
[4] Carnegie Mellon University Libraries: Digital Collections. Available at
http://diva.library.cmu.edu/ (March 2010).
[5] IEEE Standard for Learning Object Metadata. Available at
http://ltsc.ieee.org/wg12/20020612-Final-LOM-Draft.html (March 2010).
[6] Association for Computing Machinery; IEEE Computer Society.
Computer Science Curriculum 2008: An Interim Revision of CS 2001.
Available at http://www.acm.org/education/curricula/ (March 2010).
[7] ACM; AIS; AITP. IS 2002: Model Curriculum and Guidelines for
Undergraduate Degree Programs in Information Systems. Available at
http://www.acm.org/education/ (March 2010).
[8] ZLATAREVA, N; NISHEVA, M. Alignment of Heterogeneous
Ontologies: A Practical Approach to Testing for Similarities and
Discrepancies. In: D. Wilson, H. Chad Lane (Eds.), Proceedings of the
Twenty-First International Florida Artificial Intelligence Research
Society Conference, AAAI Press, Menlo Park, California, 2008, pp. 365-
370.
[9] PAVLOV, P; NISHEVA-PAVLOVA, M. Knowledge-based Search in
Collections of Digitized Manuscripts: First Results. In: Proceedings of
the 10th ICCC International Conference on Electronic Publishing, FOI-
Commerce, Sofia, 2006, pp. 27-35.

Proceedings ELPUB2010 – Conference on Electronic Publishing

June 2010 – Helsinki, Finland