T. Kerkiri, A. Manitsaris and I. Mavridis
Dept. of Applied Informatics University of Macedonia
156 Egnatia Street, 54006, Thessaloniki, Greece
Keywords: e-Learning, e-Reputation metadata, KAON, Semantic Web.
Abstract: In this paper the architecture of a learner-centered e-Learning system, which aims to qualify and recom-
mend learning material using semantic web technologies, is presented. In the proposed approach, the learner
has a central role in the decision-making process of the distribution of learning material. The main idea re-
lies on gathering appropriate reputation metadata from the learning material consumers. As a result of a
proper material ranking, a better matching to future searches of learners having similar profile is anticipated.
An experimental implementation of the presented architecture provided interesting results that better depict
the applicability and usefulness of the proposals.
The Web today is oriented to the semantic technolo-
gies which intend to make data-search meaningful,
more accurate, and detectable by machines. In se-
mantic technologies the well-formed metadata is
mainly used to express semantics and the material-
retrieval is based on it. One of the fields, among
others, that adopt the emerging technologies of the
Semantic Web is the e-Learning area. Generally
speaking, e-Learning systems are mainly attended by
adults (Rogers 1999), so they have to provide per-
sonalized knowledge, dedicated to their abilities,
skills, and knowledge demands (Brusislovski 2003).
Moreover, these systems have to promote co-
operation among their attendants in order to facili-
tate the learning procedure, to overcome problems
that arise from distance, and to elevate the learners’
personalities. Today, in Internet applications, the
elevation of the users’ role and the promotion of
collaboration among them is a major trend. This is
more and more obvious in modern e-Marketplaces
(e.g. Amazon, e-Bay, etc) which yearn for users’
participation and, at the same time, they take advan-
tage of their opinions. In the same idea, the blogs
and the social networking sites are con-
m, www.hi5.com, etc). Our work is based on the
idea that the role of the attendants is of great impor-
tance in e-Learning systems, too. Their opinions,
expressed as evaluation metadata, can improve the
functionality of the system. According to our ap-
proach, the metadata in e-Learning systems is not
static, as it is in common nowadays. It is not only
defined by the material-provider during the initiali-
zation phase of the system, but it can change radi-
cally while the system is running. Such continuously
provided accurate and countable e-Reputation meta-
data (Dellarocas et al. 2002) and (Kerkiri et al.
2006), is collected and exploited to improve the
documents’ distribution, the learning resources’
quality, and to guide future searches of learners hav-
ing similar profile.
This paper is structured as follows: in section 2,
related work on the semantic web ontological tools
is presented. In section 3 a proposed modular archi-
tecture of a learner-centered system is presented. An
experimental implementation of the proposed system
architecture for an e-Learning environment is pre-
sented and evaluated in section 4. Finally, our con-
clusions and future work are discussed in the last
Web semantic systems are based on metadata which
describe their material and on ontologies which clas-
sify their content and inference from it. Generally
speaking, such systems: i) provide an editor to create
the ontology, expressed in a standard ontological
Kerkiri T., Manitsaris A. and Mavridis I. (2007).
In Proceedings of the Third International Conference on Web Information Systems and Technologies - Society, e-Business and e-Government /
e-Learning, pages 563-566
DOI: 10.5220/0001291205630566
language (e.g. RDF, OWL), and to manipulate its
instances, ii) support a standardized method (e.g.
RDQL, RQL, RDF-QEL, SeRQL, etc) for searching
into the repositories and for sharing their contents,
iii) are able to adopt newer standards (e.g. OWL).
Some of the current semantic products/tools, like
Protégé (available at http://protege.stanford.edu), are
specialized in creating-editing ontologies. Others,
like Sesame (available at http://www.opendf.org)
emphasize on the repository sharing. SHAME
(available at http://kmr.nada.kth.se/shame/wiki/Ov-
erview/Main) is focused on the creation of special-
ized RDF-schema repositories, and especially on
educational metadata schemas, like LOM (avaliable
at IEEE Learning Technology Standards Committee.
2001, IEEE LOM Working Draft 6.1. http://ltsc.ieee.
org/wg12/index.html), DublinCore, etc. KAON
(Volz et al. 2003), OntoStudio (available at
http://www.ontoprise.de) and Sesame provide a cen-
tral container for their metadata-repository, while the
Edutella educational network (Nejdl et al. 2002)
uses P2P technologies to share its material. More-
over, KAON, Sesame and OntoStudio provide a
specific database schema that simulates the RDF and
OWL framework. Some of them create an integrated
environment; others are focused just in just one ca-
pability (editing/querying) and they do the rest of the
job with plugins. All the above mentioned products
are general purpose tools that both create the ontol-
ogy for a specific application, and manipulate its
instances, as well.
We need a semantic web based system, dedi-
cated to e-Learning area that can provide specialized
services for this field.
A schematic representation of the proposed architec-
ture, in correlation to the three phases of a method-
ology that aims to guide the implementation of a
semantically web based knowledge management
system (Kerkiri et al. 2004), is depicted in figure 1.
The architecture is designed in a modular manner
to undertake the advantages of such approach. The
architecture is based on a central RDF-based server
which collects the metadata of all e-Learning system
participants. The central repository lets the learners
to take advantage of the system’s dynamically in-
coming up-to-date metadata. Each node that wants
to find material using this system is given access on
the RDF-Server. Every annotation which is inserted
on the central repository stands as an advertisement
for newly inserted learning resources. If a participant
wants to contribute as a learning-resource provider
as well, a local material repository must be created.
On each of these material-repositories autonomous
local sharing and security policies can be applied.
According to this view the system stands as a con-
tinuously updated encyclopaedia of the knowledge-
domain defined by the ontology.
Figure 1: The modules of the proposed architecture.
The proposed architecture consists of three main
1. Knowledge creation module: it creates/edits
the ontology and handles its instances during the
knowledge creation phase. It separates the interfaces
of the learner and the learning resources’ provider
and makes them adaptable to their needs. It also de-
termines the different permissions that providers and
consumers both have upon the documents. It con-
sists of two sub-modules:
1.1. Οntology handling sub-module: it is imple-
mented in three distinct sub-systems:
1.1.1 General-annotation sub-module: it is a
general purpose RDF-editor.
1.1.2 Learning-material handling sub-module: it
provides a unique id (URI) to each learning re-
source, so that the resource can be attached the
proper properties, according to its definition.
1.1.3 Learner-handling module: it creates the
learner’s profile, and defines the access policy and
the authorizations he has on the system.
1.2 Correlation sub-module: it creates correla-
tions among the documents and their URIs. During
the retrieval phase it undertakes the applying of each
node’s policy upon the documents it distributes.
2. Search and retrieval module: it accepts the
learners’ criteria, translates them to suitable RDF-
expressions and looks for appropriate learning re-
sources. This module allows searching even into
sub-concepts of the ontology. It consists of two sub-
WEBIST 2007 - International Conference on Web Information Systems and Technologies
2.1 Search sub-module: this module i) applies
the suitable verifications regarding the learner’s ca-
pability of searching for learning material and ii)
transforms every RDF-triple (object, non-attribute
property, literal value) to an appropriate searching
2.2 Retrieval sub-module: it undertakes the con-
trol over the shared learning resources of the node.
Using the URI metadata the actual position of the
learning material is found into the network; then the
learner’s authorizations’ are checked, and, if permit-
ted, the document is revealed to the Learner.
3. e-Reputation handling module: it lets author-
ized learners to evaluate the documents, according:
i) to their context, using standard evaluation vari-
ables and ii) to the metadata they have obtained by
their providers. Three sub-modules implement the e-
Reputation metadata process:
3.1 e-Reputation collection module: it deals with
problems commonly appeared in Reputation-
systems: identity checking and e-Reputation meta-
data providers’ privacy. After the checking, e-
Reputation metadata is collected in cooperation to
the annotation sub-module.
3.2 e-Reputation metadata processing sub-
module: it creates sub-sets of i) the learners accord-
ing to the metadata included in their profile, and ii)
the learning objects, according to both their annota-
tions and their reputations.
3.3 e-Reputation inference sub-module: this
module i) uses the results of the e-Reputation meta-
data processing sub-module in order to propose to
future learners, having similar profile, documents
that could be suitable for them ii) uses the criteria
that the learner has stated during the learning mate-
rial search and combines them to their profile.
To experience the proposed architecture, an applica-
tion has been developed. The actual implementation
of the central knowledge-base repository is based on
the structure of a database system borrowed from
KAON (Motik et al. 2002), which simulates the
RDF. Using the ontology handling sub-module of
the above mentioned application, an ontology was
created, based on LOM standard. The main concepts
of an e-Learning system according to LOM are: the
LearningResource”, having properties that facilitate
its retrieval, and the “Learner”, having properties
(his profile) that facilitate the system to find the ap-
propriate learning resources for him. The ontology
classifies the material which supports the “Multime-
dia” course for the students of our educational insti-
tute. The instances of this ontology were correlated
to learning material that was distributed in 10 con-
tributed nodes, using the correlation sub-module,
providing 183 different learning resources. The aim
of this implementation was to provide suitable re-
sources to the learners needs, matching the proper-
ties of these two concepts, during the knowledge
retrieval phase.
Apart from these standard LOM entries, two new
concepts were introduced in the ontology:
i) e-Reputation concept: the instances of this
concept intend to evaluate each of the learning re-
sources metadata that has been provided from the
learning resource providers. Properties of this con-
cept may be evaluations for anyone of the criteria
that are used to classify/retrieve the learning re-
ii) EvaluationCriteria concept: instances of this
concept are criteria like: “Usability”, “Originality”,
“Comprehensitivity”, “Scientificity”, etc, that anno-
tate the content of the learning resource. Techniques
and variables used to evaluate user-centered adapt-
able systems can be found at a study of Gena (2005).
51 users, having different permissions, were
given access to the system to implement the knowl-
edge creation phase. According to their permissions,
the users were able to i) create annotations, ii) pro-
vide resources, and iii) make reputations. For dem-
onstration purposes, 452 different metadata-
searching properties, divided in 11 categories (cate-
gories 1-11, depicted in figure 2), were inserted into
the central data repository. Two more properties (12-
13) were introduced to record the satisfaction of the
system usage and its provided results.
Figure 2: The 15 categories of criteria. Statistics before
and after the 1
cycle of the system operation are de-
During the search and retrieval phase: i) many
queries were posed and their searching-criteria were
recorded, ii) the participants provided their evalua-
tions in a scale from 0 to 1, (evaluation collection
sub-module) using the previous mentioned e-
Reputation-Evaluation criteria. 561 different reputa-
tion metadata have been provided. ii) After experi-
enced the system, the learners provided their own
material in 27 cases. In 88 cases, and for a variety of
these criteria, the average was less than 0,4. In all
these cases the providers revisited their material and
changed their metadata (15,68%) to better match to
the learners’ proposed evaluations.
During the e-Reputation-Feedback exploitation
phase, the system was asked to find resources by
combining the learners’ profile metadata, as well as
the properties of the learning resources and the e-
Reputations other learners had provided. The e-
Reputation processing sub-module capabilities were
used to propose material according to this metadata
combination. According to a study of Kerkiri
(2006), the mediator used to exploit the e-Reputation
metadata was a SQLServer-2005 view that provided
the accumulative evaluation of any identical e-
Reputation criterion. More over, suitable stored pro-
cedures were created to inference from the view’s
contents. During this phase new reputations we re
provided. The average mean of the next-step reputa-
tions augmented, for each of the criteria, by a means
of 0.3 points, according to the initial average (fig.2).
Two more criteria (14-15) were added after the first
evaluation phase, to find out if the learners were
motivated to participate.
As interesting consequences of this implementa-
tion can be considered the following facts: i) the
ordering of the resources was changed, according to
the ranking they gathered, ii) less results were re-
trieved in each query, iii) better matching of the
learning resources to the “suitability” criterion was
In this paper a modular architecture, based on educa-
tional standards and Semantic Web technologies,
which aims to share knowledge over an e-Leaning
network is presented. The knowledge-consumer of a
system that conforms to the proposed architecture
has a central role in the overall functionality. Each
learner can participate to this system according to
his permissions by creating annotations, providing
his own learning resources, or/and providing his
countable e-Reputation metadata. The e-Reputation
metadata is of great significant in this architecture
and it is exploited to make the learning resources’
metadata more accurate, to improve the quality of
the learning resources context, to recommend mate-
rial suitable to each learner’s profile, and to promote
co-operation among learners.
To demonstrate the advantages of the proposals,
an experimental implementation of an e-Learning
system, has been provided. As depicted in the previ-
ous section, the system gradually improves its re-
sults on providing personalized resources, after hav-
ing collected a great amount of e-Reputation meta-
In our future plans is to develop the complete
ranking process of the learning resources and to im-
prove the functionality of the e-Reputation inference
Rogers, A., 1999. Teaching Adults, Open University
Press, ISBN 960-375-015-8
Brusislovski, P., 2003. Developing adaptive educational
hypermedia systems: From design models to authoring
tools. In T. Murray. S. Blessing, & S. Ainsworth Eds.
Authoring tools for advanced technology learning en-
Volz, R., Staab, S., Oberle, D., Motik. B., 2003. KAON
SERVER - A Semantic Web Management System,
Twelfth International WWW conference, Budapest,
Hungary, ACM
Nejdl, W., Wolf, B., Staab, S., Tane, J. 2002. Edutella:
Searching and annotating resources within an rdf-
based P2P network. Proceedings on the www2002 In-
ternational Workshop on the Semantic Web, Hawaii.
Volume 55 of CEUR-WS.org
Dellarocas, Chr., Resnick, P., 2003. Online reputation
mechanisms: A roadmap for future research
Kerkiri, T., Manitsaris, A., Mavridis, I., 2006. E-
Reputation Ontology for Adaptation in E-Learning, In-
ternational Workshop on Applications of Semantic
Web Technologies for E-Learning SW-EL, In con-
junction with International Conference on Adaptive
Hyper media and Adaptive Web-Based Systems
AH’06, Dublin, Ireland
Gena, Cr. 2005. Methods and techniques for the evalua-
tion of user-adaptive systems, The Knowledge Engi-
neering Review, Vol. 20:1, 1–37, Cambridge Univer-
sity Press
Motik, B., Maedche, A., Volz, R., 2002. A Conceptual
Modelling Approach for Semantics-Driven Enterprise
Applications, On the Move to Meaningful Internet
Systems, DOA/CoopIS/ODBASE 2002 Confederated
International Conferences DOA, CoopIS and OD-
BASE, pp:1082–1099, ISBN:3-540-00106-9,
Kerkiri, T., Manitsaris, A., Mavridis, I., 2004. Defining
adaptive e-Learning courses in Semantic Web, WSEAS
Transactions on Info science and applications Issue 1,
Volume 1, pp. 298-303, ISSN 1790-0832
WEBIST 2007 - International Conference on Web Information Systems and Technologies