Techniques, Technologies and Patents Related to Intelligent Educational Systems - E-Infrastructures and Technologies for Lifelong Learning - page 13

sification' refers to building a model that correctly
classifies data items into a number of predefined
classes (Frias-Martinez et al., 2005). The task of
'clustering' is to structure a given set of unclassi-
fied instances by creating concepts/classes, based
on similarities found on the training data (Frias-
Martinez et al., 2005). 'Generalization' refers to
the ability of classifying correctly instances not
presented during training or building of the model.
Certain AI techniques involve representation of
structural and relational knowledge. Such schemes
represent knowledge in the form of a graph (or a
hierarchy) and can be used for domain knowledge
representation. They involve, among others, se-
mantic networks , frames , conceptual graphs and
ontologies . Nodes in a semantic network graph
represent concepts and edges represent relations
among nodes. Nodes in a frame hierarchy have
internal structure describing corresponding con-
cepts via a set of attributes. Conceptual graphs are
based on semantic networks. An ontology provides
a shared vocabulary, which can be used to model a
domain (Staab & Studer, 2004). Typical ontology
components include individuals (e.g. instances/
objects), classes, individual and class attributes,
relationships concerning classes and individuals,
events, restrictions, rules and axioms. Ontology
languages (e.g. OIL, DAML, DAML+OIL and
OWL) are used to encode ontologies. Most of
these languages have been developed for the
Semantic Web. Ontologies can be exploited to
provide different levels of metadata for learning
material: they can be exploited to describe the
content (semantics), to define learning context
and to define relational and structural knowledge
involving learning material (Jovanovic, Gasevic
& Devedzic, 2006). Ontologies can play an im-
portant role in several advanced tasks. They can
be exploited for automatic creation of metadata
for learning material (Jovanovic, Gasevic & De-
vedzic, 2006), management of (collaborative)
learning material authoring describing interactions
between author(s) and learning material as well
as relationships among various learning material

versions (Nesic, Gasevic & Jazayeri, 2008), de-
scription of heterogeneous, distributed Web-based
learning resources.
Rule-based reasoning (RBR) , is a popular
KR&R method (Ligeza, 2006). Rules represent
general domain knowledge in the form of if-
then rules: if <conditions> then <conclusion>. A
conclusion is derived when the logical function
connecting its conditions results to true. Conclu-
sions are drawn from rules and known facts about
the problem at hand. Explanations about drawn
conclusions can be provided. Rules can be useful
in the following situations: (a) when they are avail-
able or easily obtainable from experts or data, (b)
when naturalness and provision of explanations is
a requirement, (c) when classification tasks need
to be performed and all input values are known, (d)
when general knowledge is necessary. Rules are
often used in most pedagogical tasks (Brusilovsky
& Vassileva, 2003; Lanzilotti & Roselli, 2007).
Certainty factor rules can be also employed to
handle uncertainty.
Case-based reasoning (CBR) stores a large set
of past cases in the case base (Kolodner, 1993).
When handling a new case, CBR retrieves the most
relevant stored case and adapts it appropriately.
Useful knowledge gained when handling the new
case is retained enhancing reasoning capabilities of
CBR system and enabling IES self-improvement.
CBR can be useful in the following circumstances:
(a) an adequate amount of past cases involving
learning activities/experiences is available/obtain-
able and their adaptation is useful for handling
similar new learning activities, (b) it is difficult
to acquire formal instructional knowledge, (c)
empirical (i.e. practical) knowledge is required to
teach the specific domain. CBR has been used for
instructional tasks such as modeling of teaching
strategies (Soh & Blank, 2008), adapting learning
contents according to learning styles, emotions
and individual needs (Sarrafzadeh et al., 2008),
teaching how to solve problems based on previ-
ous similar ones, instruction involving construc-
tivism. CBR may also be useful for modeling