See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267680371

Ontologies for eLearning

Article · January 2008

CITATION READS
1 399

2 authors:

Bartlomiej Polakowski Krzysztof Amborski

1 PUBLICATION 1 CITATION
Warsaw University of Technology
17 PUBLICATIONS 30 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Krzysztof Amborski on 22 February 2015.

The user has requested enhancement of the downloaded file.

 Ontologies for eLearning
B. Polakowski, K. Amborski

Institute of Control and Industrial Electronics, Warsaw University of Technology.

Warsaw, Poland
e-mail: polakpl@yahoo.com, ambor@isep.pw.edu.pl

Abstract
Learning is one of these activities that differ from one person to another. However most of the
eLearning systems do not include any mechanisms that could support pedagogical diversity.
The individual aspects of the stu dents and their needs are ignored what reflects in worse
quality of learning. In this paper, we present the architecture and features of the present
eLearning systems and show the newest trends in this field, directed on dynamic knowledge
acquisition, and personalization of presentation. The key role in this solution is use of
ontologies as methods for knowledge representation. We present how an ontology can be used
in a learning environment in order to give the student the content that will fit his needs a nd
support him during the learning process.

Keywords
Ontology, personalization, feedback, eLearning, LMS, Web 3.0

1. Introduction to Learning Management Systems

E-learning would not be so successful without some kind of platform that would
allow control over the whole learning process, its parts and participants. In the past
the e-courses were scattered all over the Web or companies intranet (btw such
solutions can still be found). The learners were let on their own what brought nothing
except lost money. In most cases users were not interested in using these resources,
because learners could not find them and tutors could not monitor learners’ progress.
The solution came with first learning platforms.

Learning management system (LMS) or a Virtual Learning E nvironment (VLE) is a

“software system designed to help teachers by facilitating the management of
educational courses for their students, especially by helping teachers and learners
with course administration” (Wikipedia, 2008). In other words it’s a comp uter
program that facilitates a computerized learning by organizing, mapping and
delivering content and managing different types of users (i.e. learners, instructors,
administrators) and their interactions. It allows easy management of the learning
materials and tracking of user’s learning progress and results. LMS is typically a
large system that includes many features such as: access control, personalized
working spaces, assessment tools, searching facility, content creation tools,
information distributio n mechanisms (i.e. Calendar, News announcements), or
communication and collaboration functions (i.e. email, chat, discussion forum) .
We can divide them into three groups depending on the type of user and functionality
he is offered. (Culley, 2006)
Features available for all users:

 Authentication (including login, name and password)

 Individualized working place (home page) based on user’s settings and

preferences.

 Support (help pages and quick links)

 Authoring tools for different types of content

 Information tools (i.e. Calendar, News & Course announcements ).

 Synchronous and asynchronous communication and collaboration functions

(i.e. email, discussion boards, chats).

Features offered to the tutor:

 Users access control

 System management tools

 Ability to track student progress.

 Assessment tools.

 Student Assignment Management,

 Administrative applications.

Features offered to the student:

 Personalized access to learning resources

 Learning management tools.

 Online support (feedback from system or ot her users)

 Self assessment (i.e. online quizzes)

A generic LMS architecture is presented in Figure 1.

 Figure 1 Generic LMS architecture ( Wise, 2005)

A common architecture is very simple and includes four main components. A

webserver (i.e. Apache) is the central component that stores content repository
(course web pages) and the whole application layer including all mentioned before
mechanisms such as authentication or communication tools. All data about users,
system configuration, and information about available courses is stored in database
(typically MySQL) in database layer. There is no need to install any special client
software other than standard browser, because learning content can be provided in
any web-deliverable format.

One of the most popular and the best Learning Management System - Moodle
(Moodle, 2007) is based on such architecture. A study called “An Evaluation of
Open Source E-Learning Platforms Stressing Adaptation Issues” found that Moodle
outperforms all other platforms and also obtained the best rating in the adaptation
category. It is a modular object -oriented dynamic learning environment which can
be customized to any need with new ready or self written modules.

In large companies LMS is often integrate d with other large systems such as Content
Management or Human Resources. All subsystems share an interface in a form of
custom portal. Because of the large number of users there must be also LDAP
Directory service. Figure 2 presents enterprise architectur e with LMS included.
 Figure 2 Enterprise architecture ( Wise, 2005)

Such solutions have been already successfully implemented at thousands of

universities and companies. However present learning management systems should
follow the newest trends in information technologies and e -learning in order to offer
more effective methods of learning. Web 2.0 brought more creativity, collaboration,
and sharing possibilities to the users. Such things as blogs, wikis, podcasts, RSS
feeds , and social software are to find in all professional LMSs. These technologies
gave learners possibilities to become active participants in the learning process. They
are no longer passive “receivers” of knowledge, but actively take part in their
creation.

The time has come for another big step. The next generation of World Wide Web
called Web 3.0 has already begun. Web 3.0 technologies, such as intelligent software
that utilize semantic data, have been already implemented and used by multiple
companies. It was caused by the growing interest to bring semantic web technologies
to the general public. E -learning authoring tools vendors have already given
possibilities to create metadata for content in order to allow it’s recognition and
reasoning in incoming Web (Reload, 2007). The e-learning systems are unprepared
for such change so far. Present researches of key players in this field go in the
direction of workflow and document management systems (DMS). However the
progress of semantic Web technologies is unavoida ble. On a few universities there
are already trail versions of new e -learning systems. They exploit ontologies that in
our opinion are the future of knowledge representation. These platforms offer a new
feature to the user – personalization of content. Wit h this functionality the learning
process will become more similar to ideal one to one learning with teacher. Normally
it is unprofitable or even impossible to accurately and instantly give the best
feedback or learning materials to the user. Systems based on ontologies keep the
profile of each user and thanks to the ontologies are able to find best fitting content.
2. Introduction to Ontologies
Ontology can be described as a formal description of knowledge from some domain.
It can be understand as a whole kno wledge used in a computer system or just as a
class model of one domain. One of the most popular definitions of an ontology
comes from (Gruber, 1993) and it says “ontology is a description (like a formal
specification of a program) of the concepts and rela tionships that can exist for an
agent or a community of agents”. He used the term ontology to mean a
“specification of a conceptualization” . The weak points of this statement related to
different understanding of the same words in different countries have been pointed
out by (Bassara, 2004). However for our needs we will use the definition from
Gruber.

The growing interest in the idea of the Semantic Web (Berners -Lee et al., 2001),
reflects on ontologies which are seen as mechanism that will allow the
interoperability of web resources. The research in this field brought a standard
ontology language based on XML, Resource Description Language (RDF), RDF
Schema (RDFS), and the Web Ontology Language (OWL). (Bechhofer et al., 2004)

The eLearning community disco vered the great potential of ontologies

and the Semantic Web (Sampson et al., 2004) (Lytras et al., 2003) and started
researches that have an aim to make use this of this mechanism within eLearning.
According to the work of (Stojanovic et al., 2001) we can distinguish three different
dimensions in eLearning that can be described using ontologies: content, context
which defines the way in which the content is presented and the structure of the
learning material within a course. The most advanced researches have been made in
direction of the first group what have resulted in such systems as Edutella which is a
peer to peer network for searching semantic web metadata. (Edutella, 2007)

In our work we have concentrated on di fferent approaches to use ontologies within

eLearning environments. We have recognized main streams in this direction,
described them and showed their implementation. The first one is personalization of
content, which enables the system to respond differen tly according to the user
characteristics, performance and current knowledge. The second is improvement of
learning resources search, which will helps to get data from external repositories that
is topically connected with our current activity and properti es. Last but not least
important topics are ontologies within architecture of feedback systems which
provide support for both students during learning and teachers during authoring.

3. Using ontologies for content personalization

Personalization is a very im portant issue in the evolution of the modern eLearning
systems. There are several learner types that require different kind (in different form
i.e text, pictures and videos) of learning materials for more effective learning.
(Bleimann, 2004) Every learner has different knowledge and competences. This
forces eLearning systems to know the profile of each student , monitor his progress,
and choose always content that will exactly fit his needs. An interesting approach to
this problem was presented by (Gomes et al, 2005). It is based on a student model
which defines all data about student and is built incrementally by the system using
several data sources (i.e. student enrolment form, data from interaction with system,
teacher). The collected data can be divided into two groups: static, which is an
aggregation of student characteristics that are not changed during a learning process
(i.e. name, qualifications), and dynamic which gathers information about the
student's current performance in the course, and describ es student's knowledge and
competences relevant for the current course. The whole information is the basis for
the personalization model. (Gomes et al, 2005) distinguish two models. The off-line
model (Fig. 1) allows content developers to change course con tent according to
analysed data (using data mining tools) from student system interactions.

Figure 3 Architecture of off-line information model (Gomes et al, 2005)

The on-line model monitors interaction between student and syste m and tries to
adapt the content according to the student model. There must be ofcourse a reasoning
engine that decides what interaction data must be adapt. The architecture of thi s
solution is presented in Fig.2

Figure 4 Architecture of on-line information model (Gomes et al, 2005)

To decide what content should be presented to the student, there must be a reference
between the dynamic model and course concepts organized in an ontology. The
ontology represents the domain knowledge and is the formalization of its concepts.
In (Gomes et al, 2005) model, the taxonomic structure is built from classes which
represent the concepts and generalization relations between them. With some course
modules, objectives and interactions there are a ssociated concepts. The system can
trace their status during the session and so determine the student's knowledge. It
allows computing the learning progress of a student. Each concept in the ontology
has associated parameters and an associated state. The p arameters define the number
of correct and wrong answers as well as the number of completed SCOs by the
student. The state is based on these parameters, and it stores information if and how
good the concept was learned. The data about student is updated in the domain
ontology what enable to trace his knowledge and progress at any time.

4. Ontology based feedback generation

Feedback can be described as a process that have an aim to inform a user (in our case
a student) about the accuracy of his response to some instructions. It can inform not
only about answers correctness but also about precision, timeliness, learning
guidance, motivational messages, background materials, learning focus and others
(Mory 2003). Feedback is a crucial element in every user oriente d system. In the
eLearning environments it is particularly important because it is one of the main
issues that defines the quality of learning. In the current generation of eLearning
systems, feedback is often reduced to the email communication with a tut or, what is
very time consuming and ineffective, or to automatically produced, hard coded
support, which is often invaluable. There is one exception in environments based on
social constructivism where learners and teachers solve problems through social
negotiations (Duffy, 1996). Semantically rich feedback can be only found in a
specialized design environment .

There are already eLearning systems that are based on ontologies (Aroyo, 2002 (1)
(2)), (Jin, 1999). The eLearning community carries out researche s in order to develop
a mechanism, based on ontologies which would supply valuable feedback both
during the learning process and authoring phase. It has to be generic, domain and
task independent and produce semantically rich feedback. Such an approach was
presented by (Passier et al, 2003). They developed a feedback mechanism where
ontologies are used as arguments of the feedback engine and mechanism which
enables the authors to define a domain or task specific feedback. Both things can be
reused for different ontologies what saves a lot of time and developers work. Such a
system gives student a possibility for example to compare the completeness and
correctness of his solution with education ontology, which describes the learning task
and was created by the author. The learner can be asked to list some concepts and
their relations or choose and relate concrete instantiations. On the other hand an
author can analyse the domain ontology and get feedback about the concepts. For
example some of them can be used but never defined or they can be used before their
definition. Such system must by ready to solve problems with synonyms or
homonyms between concepts.

The system that delivers such generic feedback mechanism (see Figure 3) naturally
contains several types of knowledge which is represented using ontologies (Passier et
al, 2003).
Figure 5 Architecture of an eLearning system with generic feedback generation
(Passier et al, 2003)

The main architecture contains three main components: a player where the learner
can learn concepts, construct artefacts and solve problems, an authoring tool where
an author can develop courses and course related materials (i.e. ontologies,
examples, feedback patterns) and a feedback engine, which uses ontol ogies as
parameters and automatically generates feedback. It can be generic, that means
independent of ontologies (applicable to all design activities ) or specific, which is
defined by the author and can be task, domain or, course specific.

To construct feedback the presented engine needs five ontologies as arguments,

which can be freely exchanged (Passier et al, 2003).

5. Ontology for improving learning resource search

The common problem in the development of distributed eLearning environments is a
lack of the interoperability between the systems that are based on different
ontologies. One of the main issues is connected with the learning materials. Since a
few years there are special digital libraries (i.e ACM Portal) and community -based
learning repositories (i.e. Merlot) that allow the exchange of such materials.
However, researches showed that authors have difficulty of making annotations of
their learning material using subject taxonomies (Greenberg et al., 2002). This
reflects on lower effectivity of sea rching across multiple object collections. The
resources are not interconnected into the Web, so there is no context where they can
be defined.

The main aim of current research is to develop a mechanism that will allow users to
find resources that are mo st relevant in their current learning context. The solution
looks simple. There are already web based courses which structure can be considered
as ontology. The learning resources that form them are directly related to their
structure and annotated with th e course ontology. (Gaševic et. al, 2005) proposed a
course structure that is similar to the SCORM standard. On the left hand side of the
screen user sees a structure of the course and on the right hand side the appropriate
content. The bottom part contai ns a search mechanism that allows context sensitive
search for the ACM Digital Library (ACM DL, 2005). What makes this solution
different from traditional text search engines is that the search request together is
collected together with annotation of the current page within the course ontology.
Than it is sent to the ontology mapping system, which maps the course ontology to
the ACM CCS (classification in the domain of computer science) ontology and sends
an expended query to ACM DL. (Gaševic et. al, 2005 ) pointed out that two areas in
the context of their work that need further research.

Figure 6 Ontological levels of learning object and domain ontology (Passier et

al, 2003)

The first thing is annotation of learning objects wit h Domain Ontologies. The
ontological levels of learning object and domain ontologies are different so the
learning object metadata which is an instance of learning object metadata schema is
enriched with keywords defined in the ontology (Figure 4). In the schema there are
keywords that are defined as classes. If we refer to them, we annotate the metadata
(i.e. metadata instances) with ontology classes (i.e schema). Using classes as
concepts complicates reasoning, but there is a solution in a form of special ized
ontologies for defining domain taxonomies with a rich set of properties for defining
concept hierarchies. Another issue is mapping between multiple domain ontologies.
There are many ontology mapping techniques. (Gaševic et. al, 2005) system uses a
mapping ontology that expresses relations between two ontologies. (Figure 5)
 Figure 7 Mapping ontology (Gaševic et. al, 2005)

The two problems mentioned above can be solved using currently developed Simple
Knowledge Organization Sys tem (SKOS) (Miles & Brickley, 2005). It allows
defining different types of ontologies (i.e. taxonomie, thesaurus) and mapping of
concepts from different domain ontologies, using special RDF vocabularies. These
vocabularies provide a model for expressing t he basic structure and content of
concept schemas, and describing mappings between concept schemes. The mappings
specify the mapping relations between concepts using set of properties like
broadMatch, narrowMatch, ExactMatch etc. which reflect weight of ma pping. This
facilitates the later ranking of search results. In the case of (Gaševic et. al, 2005), the
defined mappings enabled the user to improve searching in the remote ACM DL,
using local context in the form of course ontology. What is more, not all m appings
have to be defined, because a special algorithm allows extrapolation of missing
mappings using the structure of both ontologies. It takes concepts of the source
ontology as input arguments and concepts of the target ontology as the results of the
mapping algorithms. Than SKOS Mappings are used to define mapping relations
between concepts from both ontologies. The search results are ranked according to
different types of SKOS Mappings properties. The algorithm uses also all the
“children” (sub-nodes) of matched concepts in the target ontology. The weight
factor is used to determine ranks of matched concepts and their children in the
results. It is calculated according to the type of mapping relation connecting the
matched concept with the source onto logy concepts. The researches showed that the
use of ontologies reduces the number of found objects during the search activities,
what helped learners to find resources that better fit their current learning activities
(Gaševic et. al, 2005).

6. Conclusion
This paper presents the advantages of ontology based mechanism within modern
eLearning systems. We have presented three main directions in current researches in
this area and give concrete examples of implementation. The main aspect of our
work was to provide the reader with the idea about the current state of eLearning
technologies and to show further research area.

7. References
Culley A., (2006) Learning Management Systems ,
http://www.instructionaldesign.com.au (Accessed 11 December 2007).

Moodle, (2007), http://www.moodle.org.

Wise L., (2005), What is Online Learning?, http://onlinelearningunit.com.

Reload, (2007), http://www.reload.ac.uk/.

Gruber T. R., (1993) A translation approach to portable ontologies. Knowledge

Acquisition, 5(2), pp199-220.

Bassara A., (2004) I weź tu się dogadaj – ontologie, Gazeta IT nr 1(20).

Berners-Lee, T., Hendler, J., Lassila, O. (2001). The Semantic Web, Scientific
American, 284(5), pp34 -43.

Bechhofer, S., van Harmelen, F., Hendler, J. Horrocks, I., McGuinness, D.L., Patel -
Schneider, P.F., & Stein, L.A. (2004). OWL Web Ontology Language Reference,
W3C Recommendation, http://www.w3.org/TR/owl -ref/ (Accessed 9 December
2007).

Sampson, D. G., Lytras, M. D., Wagner, G., & Diaz, P. (2004). Guest Editorial:
Ontologies and the Semantic Web for E-learning, Educational Technology &
Society, 7 (4), pp26-28.

Lytras, M., Tsilira, A., Themistocleous, M. (2003). Towards the semantic e -learning:
An ontological oriented discussion of the new research agenda in e -learning, In
Proceedings of the 9 th Americas Conference on Information Systems, Tampa, USA,
pp. 2985-2997.

Stojanovic, Lj., Staab, S., Studer, R. (2001). eLearning based on the Semantic Web,
In Proceedings of the World Conference on the WWW and the Internet (WebNet
2001), Orlando, USA.

Edutella, (2007). http://www.edutella.org (Accessed 9 December 2007).

Bleimann, U. (2004). “Atlantis University – A New Pedagogical Approach beyond

E-Learning”. Campus-Wide Information Systems, 21(5), pp191 –195.

Gomes, P., Antunes, B., Rodrigues, L., Santos, A., Barbeir, J., Carvalho, R. (2005).
Using Ontologies for eLearning Personalization.
 Mory E., (2003). Feedback research revisited, in Handbook of research for
educational communications and technology, D.H. Jonassen (Ed.), MacMillian
Library Reference, Ne w York, USA.

Duffy, T., Cunningham, D., (1996). Constructivism: Implications for the design and
delivery of instruction, in Handbook of research for educational communications and
technology, D.H. Jonassen (Ed.), MacMillian Library Reference New York, USA.

Aroyo, L., (2002) (1). A layered approach towards domain authoring support,
International Conference on Artificial Intelligence (ICAI’02), Las Vegas, USA.

Aroyo, L., (2002) (2). Courseware authoring tasks ontology, International

Conference on Computers in Education (ICCE’02), Auckland, New Zealand.

Jin, L., Chen, W., (1999). An ontology -aware authoring tool, In Artificial
intelligence in Education, S. Lajoie and M. Vivet (Eds), IOS Press, Amsterdam, The
Netherlands.

Passier, H., Jeuring, J., (2003). Ontolo gy based feedback generation in design -
oriented e-learning systems.

Greenberg, J., and Robertson, (2002). Semantic Web Construction: An Inquiry of

Authors'Views on Collaborative Metadata Generation, In Proceedings of the
International Conference on Dublin Core and Metadata for e -Communities 2002,
Florence, Italy, pp45-52.

Gaševic, D., Hatala, M., (2005), “Ontology mappings to improve learning resource
search.”

ACM DL, (2005), Association for Computing Machinery (ACM) Digital Library,
http://www.acm.org/dl (Accessed 9 December 2007).

Miles, A., Brickley, D., (2005), SKOS Core Vocabulary Specification, W3C
Working Draft, http://www.w3.org/TR/swbp -skos-core-guide (Accessed 9 December
2007).

View publication stats