A TOOL FOR MANAGING DOMAIN KNOWLEDGE IN
INTELLIGENT TUTORING SYSTEMS
Panayiotis Kyriakou, Ioannis Hatzilygeroudis and John Garofalakis
Department of Computer Engineering & Informatics
University of Patras, Patras, Greece
Keywords: Information system, knowledge system, meta-data management, intelligent tutoring system.
Abstract: Domain knowledge (DK) is a basic part of an intelligent tutoring system (ITS). DK usually includes
information about the concepts the ITS is dealing with and the teaching material itself, which can be
considered as a set of learning objects (LOs). LOs are described by a data set called learning object
metadata. Concepts are usually organized in a network, called a concept network or map. Each concept is
associated with a number of LOs. In this paper, we present a tool for managing both types of information in
DM: creating and editing (a) a concept network and (b) learning object metadata. Additionally, the tool can
produce corresponding XML descriptions for each learning object metadata. Existing tools do not offer all
the above capabilities.
1 INTRODUCTION
Recent developments in computer-based educational
systems gave rise to a new generation of them
encompassing intelligence in order to increase their
effectiveness, called intelligent educational systems.
Intelligent Tutoring Systems (ITSs) constitute a
popular type of intelligent educational systems. ITSs
take into account the user’s knowledge level and
skills and adapt presentation of the teaching material
to the needs and abilities of individual users. This is
achieved by using Artificial Intelligence techniques
to represent pedagogical decisions as well as domain
knowledge and information regarding each student.
ITSs were usually developed as stand-alone systems.
However, the emergence of the WWW gave rise to
Web-based ITSs (Brusilovski, 1999;
Hatzilygeroudis, 2004).
The structure of an ITS is illustrated in Figure 1.
An ITS consists of three main modules: (a) the
domain knowledge, which contains the teaching
content and meta-information about the subject to be
taught, (b) the user model, which records
information concerning the user, and (c) the
pedagogical model, which encompasses knowledge
regarding various pedagogical decisions.
In the domain knowledge, the teaching material
must be structured in such a way that can be easily
recognized and used by the pedagogical unit, in
order to adapt teaching to user’s needs. A quite
helpful way is to distinguish between the teaching
material itself and its meta-information, typically
called its metadata. The teaching material itself
usually consists of learning objects (LOs), which are
autonomous, self-contained digital entities (e.g. web
pages) used to support learning (Wikipedia, 2008).
To be able to manage domain knowledge in an ITS,
we need a tool that will be able to manage LOs
metadata and real teaching material too. However,
apart from those, domain knowledge also contains
information about the concepts the system is
concerned with. Although there are a number of
tools dealing with management of metadata for LOs,
most of them are not suitable for dealing with
concepts. On the other hand, although there are tools
dealing with concepts, they are not able to relate
them with LOs.
In this paper, we present a tool that is able to deal
with both LOs and their metadata description as well
as with concepts and their relations. Also, it helps
tutors in preparing lessons. Description of LOs
metadata is based on the IEEE LOM standard (IEEE
LTSC, 2002; Holzinger et all, 2001).
5
Kyriakou P., Hatzilygeroudis I. and Garofalakis J. (2008).
A TOOL FOR MANAGING DOMAIN KNOWLEDGE IN INTELLIGENT TUTORING SYSTEMS.
In Proceedings of the Third International Conference on Software and Data Technologies - ISDM/ABF, pages 5-11
DOI: 10.5220/0001872100050011
Copyright
c
SciTePress
Figure 1: The Basic Structure of an Intelligent Tutoring
System.
The paper is structured as follows. Section 2
presents the structure of domain knowledge in an
ITS and its requirements of managing it. Section 3
deals with the functional characteristics of the
introduced tool, while Section 4 presents some
design and implementation issues. Section 5 presents
related work and shows the inadequacies of existing
tools. Finally, Section 5 concludes the paper.
2 DOMAIN KNOWLEDGE
The motivation for creating such a tool was the need
for managing the domain knowledge of a certain
ITS. To be able to manage its domain knowledge, it
has been structured as displayed in Figure 2. So, it
consists of three components: knowledge concepts,
course units and meta-description.
The knowledge concepts are elementary pieces
of information of the specific domain. Every
concept has a number of general characteristics such
as the name, the difficulty level, the detail level, the
prerequisite knowledge, etc. Moreover, a concept
has relations with the other concepts which mainly
show the prerequisite concepts that contain the
perquisite knowledge for that concept. For example,
in teaching logic in an Artificial Intelligence course,
‘logic syntax’ and ‘logic semantics’ could be two
concepts. ‘logic syntax’ could have as prerequisites
the concepts ‘constant’, ‘variable’, function’ etc.
However, there may be other types of relations, like
“generalizes”, “specializes”, “part-of” etc. For
example, ‘left-hand expression’ may be “part-of”
‘implies expression’. The concepts and their links
form a network (see Figure 3), which is a semantic
network that represents the pedagogical structure of
the teaching subject.
The teaching material consists of two parts: (a)
the course units and (b) the meta-description. The
course units mainly are in the form of web pages and
are equivalent to LOs. Each course unit may contain
a variety of data types (e.g. text, images, animations
etc).
Course units are used in composing lessons.
Each course unit is related to a concept and through
the concept network, that has been created, the
system chooses the next course unit (web page) to be
presented to the user. A course unit can be of a
theory, an example or an exercise type. Examples
help the student to understand in a better way the
theory. Exercises are based on the examples and are
used to evaluate the knowledge level of the user.
This information is used to update the model of the
user.
Figure 2: The Structure of Domain Knowledge.
Figure 3: A Concept Network.
The knowledge domain also contains meta-
descriptions of the course units and their main
attributes. Such attributes are mainly its difficulty
level, its pedagogical type (theory, example,
exercise), its representation type (text, image,
animation), its detail level, etc. Those meta-
Domain
Knowledge
Pedagogical
Unit
User
Mdli
User Interface
Knowledge
Concepts
Meta-
description
Course Units
ICSOFT 2008 - International Conference on Software and Data Technologies
6
descriptions of the course units are based on the
IEEE LOM standard schema (IEEE LTSC, 2002).
3 A TOOL FOR MANAGING
DOMAIN KNOWLEDGE
We created a tool for managing the domain
knowledge of an ITS based on the above. So, using
the tool one can create the concept network of a
domain, using a variety of relation links, which
him/herself can define. Also, can create, store, view
and edit the metadata for the LOs (i.e. course units)
in XML format.
3.1 Concept Map
The tool provides facilities for easily creating a map
(network) of the concepts involved in the domain.
Figure 5: New Concept Creation.
To create a new concept, one can click on the tool
workspace and choose “Add concept”. Then the
interactive form of Fig. 5 appears where he/she can
define the name of the new concept.
To create a relation between two concepts, one
should click on the
icon, lying at the right-hand
side of a concept (Fig. 7), and then on the related
concept. The interactive form of Figure 6 appears
where he/she can choose a relation type from a list
of existing ones or create a new type of relation.
Figure 7 shows two concepts connected with a
“requires” relation. At each concept one can see, at
the first glance, the name of the concept and its
relations with other concepts. Also, when the mouse
is over a concept one can see more icons that
perform certain functions. The
icon calls the
renaming function; the
icon calls the function
that deletes a relation; and the
icon deletes the
selected concept.
Figure 6: Form for adding a relation.
The user can insert concepts at any position of the
concept map, drag the concepts across the map in
order to organize them, create new relations and
connect them. In Figure 8, a concept network, where
concepts are connected with various types of
relations, is presented.
Figure 7: Concept Map Creation.
For the convenience of the users, there is a search
bar, which we can give in any keyword related to
concepts or to the metadata of the learning objects.
The result is to center the concept map to the
selected concept and highlight it.
3.2 Managing Learning Objects
The learning objects (course units) attached (i.e.
related) to each concept can be displayed by double
clicking on the corresponding concept (see Fig. 9).
We can add/create a learning object, by clicking on
the “Add Learning object” link or delete any of the
existing ones (not shown in Fig. 9).
Having displayed the learning objects, we can do
two other things:
(a) Display and edit its metadata, by clicking on
the “Edit Metadata” link of a learning object
(see Figure 9). Its metadata appears in a
structured way (see Figure 10) so that can be
easily read and edited (i.e. modified/deleted),
using the corresponding link in each metadata
category. The structure of metadata follows
the IEEE LOM standard. On this set of
metadata there is the ‘Relation’ category,
which however is not the same as the
‘relation’ between concepts. It specifies
relationships between the learning objects
(course units).
A TOOL FOR MANAGING DOMAIN KNOWLEDGE IN INTELLIGENT TUTORING SYSTEMS
7
Figure 8: A Concept Map (Network).
Figure 9: Displaying Learning Objects.
(b) Create an XML file representing the object’s
metadata. The information that we have
specified in the concept metadata is stored in
our database as well as in XML files for
future use. These XML files can be accessed
by other elements of any other system that
has access to our file system, which through
that can decide on the selection of the
concepts that will use in a tutoring session.
Figure 10: Displaying and Editing Metadata of a Learning
Object.
Therefore, we have included access to these
files through links of our system. To see the
XML file of a LO we should just click on the
“Create XML file” link of the LO (Fig. 9).
The XML file is then dynamically created
(Fig. 11).
ICSOFT 2008 - International Conference on Software and Data Technologies
8
Figure 11: The XML file of a Learning Object.
4 SYSTEM DESIGN AND
IMPLEMEMNTATION
The heart of the system is the relational database,
which is designed to store all vital information
concerning the concept description and connections,
the LOs metadata as well user related data. The
database is organized in 17 tables, apart from the
one related to ‘learning object’ (see Fig. 12), 11 for
metadata, 5 for the concept map and 1 for user
identification.
WAMP5 version 1.6.1, containing Apache 1.3.,
PHP5 and MySQL4, was used for the
implementation of the system. MySQL is used for
the database, whereas PHP for the rest of the system,
like e.g. the creation of XML files. For the concept
map drawing, javascript was used from the Open-
JACOB Draw2D library and the scipt.aculo.us
library. The queries to MySQL are made through
AJAX technology that connects the index.html file
with the php files.
5 RELATED WORK
We distinguish two categories of tools to compare
with ours. There are a number of tools, which are
used for the management of metadata in learning
objects. In spite of their many features, none of them
was satisfactory to be used with the ITS.
LOM Editor (LOM-Editor, 2001) developed by
Darmstadt University of Technology -Germany, and
written in Java, is a standalone desktop application
which includes superior abilities for editing
metadata, such as: Tabular presentation of metadata
categories; vocabulary management; multiple-
language values management; metadata template
generation to avoid the necessity to repeatedly enter
the same data in multiple fields. Some of the
drawbacks of this metadata authoring tool include:
No help or documentation; omission of specific
details of LOM Model (e.g. multiple-language
values support for metadata elements); some
standards like vCard and ISO 8601 DateTime
standard are not supported in the representation of
the metadata elements; storage of the metadata
record is done in a database and there is no export
option for the XML document.
ALOHA II (ALOHA II, 2004) is a Java-based
tool that is used for indexing, aggregating, sharing,
multi-purposing, and re-purposing learning objects.
It is created to meet the needs of indexers, educators
and learners and includes versatile and powerful
indexing tools and flexible searching of multiple
educational object repositories. The software is
based on the educational standards of IMS and
SCORM. ALOMA II is not web-based and not
based on the IEEE LOM standard.
Curriculum Online Tagging Tool (Curriculum
Online Tagging Tool, 2008) which is designed to
make the process of creating metadata and
outputting as easy and intuitive as possible. A
version of this tool is being developed to support UK
LOM Core and aspects of CanCore. It enables
creation and storage of details about the learning
resources. It also allows adding those details to the
Curriculum Online portal, so that teachers can find
out about the learning resources. Once the details
about a resource are added to the portal, the tagging
tool can be used to update them at any time, or even
remove them completely. This tool cannot create a
network of resources, but just an unstructured
repository of them.
Explor@-II (EXPLORA-II, 2008) is a software
environment for the delivery of courses or distance
learning events on the Internet. It allows creating a
virtual training centre that delivers a set of courses
on the Internet according to a variety of models and
using a LO repository facilitating information
access, production, follow-up and coaching of
learners as well as training management. It is fully
compatible with the IEEE LOM, Cancore and
Normetic. Explor@-II is more an e-learning
environment, not web-based, rather than a domain
knowledge management tool.
A TOOL FOR MANAGING DOMAIN KNOWLEDGE IN INTELLIGENT TUTORING SYSTEMS
9
Figure 12: Database Relations Diagram.
The second category concerns tools that mainly
deal with creating concepts and concept maps.
Inspiration (Inspiration, 2008) uses a web as a basic
graph structure, and repositions the initially entered
concept in the middle of the screen. This tool does
not enforce any particular graph structure, and the
representation does not require linking phrases. The
software allows the user to switch from graph to
outline view and back. SMART Ideas (SMART
Ideas, 2008) allows users to create multi-level
Concept Maps to organize ideas, to link Concept
Maps to files and Web sites, to switch between
diagram and outline views, and to publish Concept
Maps on the Web. LifeMap (LifeMap, 2008) is
designed for free educational use. Group packages
with support are available. LifeMap provides the
capability to make Vee diagrams as described in
(Novak and Gowin, 1984). IHMC CmapTools
software kit (Cañas et al, 2003a) is to enable users to
collaborate during Concept Map construction and to
easily share and publish the resulting knowledge
models. The software is based on a client-server
architecture (Cañas et al, 2003b) that allows users to
share and browse Concept Maps stored in
CmapServers distributed throughout a network that
covers the whole world. Luckie Concept Connector
is a software suite currently in development at
Michigan State University. This system allows
ICSOFT 2008 - International Conference on Software and Data Technologies
10
students to build Concept Maps online and to
receive immediate feedback about their maps based
on automatic scoring systems that are derived from
scoring methods detailed in (Novak and Gowin,
1984). The Concept Mapping system is based upon a
pre-defined set of concepts and linking phrases. The
system is currently being used for online homework
assignments.
6 CONCLUSIONS
In this paper, we present a tool that is suitable for
managing the domain knowledge in an ITS. More
specifically, the tool allows for creating, viewing,
editing and deleting knowledge concepts, which are
organized in a network (map). Concepts are
connected between each other by a variety of
relations (e.g. “requires”, “isa” etc), which can be
user defined. Also, it allows associating concepts
with corresponding learning objects, i.e. real
teaching material. In addition, the user has the
capability of adding to or modifying metadata
(which complies with IEEE LOM data model) of
each learning object associated with a concept and
create XML files which he/she can view and edit in
the process.
All the above are not offered by existing tools, as
reported in Section 5. We’ve used the tool for the
creation of the domain knowledge of an ITS
teaching “Intelligent Tutoring Systems” (see Figure
8) and another one teaching “Aspects of Artificial
Intelligence”. It has been really very helpful.
REFERENCES
ALOHA II, 2004. http://aloha.netera.ca/about.php
Brusilovski P., 1999. Adaptive and Intelligent
Technologies for Web-based Education. In C.
Rollinger and C. Peylo (Eds), Kustliche Intelligenz,
Special Issue on Intelligent Systems and Teleteaching,
4, 19-25.
Cañas, A. J., Hill, G., Carff, R. and Suri, N., 2003a.
CmapTools: A knowledge modeling and sharing
toolkit, Technical Report IHMC CmapTools 93-01,
Institute for Human and Machine Cognition.
Cañas, A. J., G. Hill, A. Granados, J. D. Pérez and C.
Pérez, 2003b. The network architecture of CmapTools,
Technical Report IHMC CmapTools 93-02, Institute
for Human and Machine Cognition.
Curriculum Online Tagging Tool, 2008. http://www.
curriculumonline.gov.uk/SupplierCentre/
taggingtool.htm. Accessed: March, 2008.
EXPLORA-II, 2008. http://explora2.licef.teluq.uquebec.
ca/ demo/. Accessed: March, 2008.
Hatzilygeroudis I., Guest Editor. Special Issue on AI
Techniques in Web-Based Educational Systems.
International Journal on AI Tools (IJAIT), 13(2),
2004.
IEEE LTSC, 2002. Draft Standard for Learning Objects
Metadata. http://ltsc.ieee.org/wg12/files/
LOM_1484_12_1_v1_Final_Draft.pdf .
Holzinger, A., Kleinberger, Th. and Müller, P., 2001.
Multimedia Learning Systems based on IEEE
Learning Object Metadata (LOM), Proceedings of
ED-MEDIA 2001 World Conference on Educational
Multimedia, Hypermedia & Telecommunications,
Tampere, Finland, June 25-30.
Inspiration, 2008: http://www.inspiration.com/home. cfm.
Accessed: March, 2008.
LifeMap, 2008. http://www.robertabrams.net/
conceptmap/lifemaphome.html. Accessed: March,
2008.
LOM-Editor, 2001. Darmstadt University of Technology:
http://www.multibook.de/lom/en/ index.html.
Novak, J. D. & Gowin, D. B., 1984. Learning how to
learn. New York: Cambridge University Press.
SMART Ideas, 2008. http://www.smarttech.
com/products/smartideas/index.asp. Accessed: March,
2008.
Wikipedia, 2008. http://en.wikipedia.org/wiki/
Learning_object_metadata
A TOOL FOR MANAGING DOMAIN KNOWLEDGE IN INTELLIGENT TUTORING SYSTEMS
11