THE ROLE OF LEARNING OBJECT ONTOLOGIES

Juha Puustjärvi

Lappeenranta University of Technology. Skinnarilankatu 34. Lappeenranta. Finland

Keywords: Learning object, Ontologies, Metadata, Learning object content models.

Abstract: Educational information systems should provide easy access to learning material. In addition learning

material should be linked in a way that the learner can easily access all required material. Unfortunately

current metadata standards and learning object content models do not support such features as they do not

capture enough semantics of learning objects. In this article we present the ontologies that give the

semantics for metadata descriptions, and thus support semantic querying and conceptual navigation of

learning objects. Semantic querying differs from traditional keyword based searching in that searching

expressions are based on ontologies, which describe the concepts of the domain in which learning takes

place. Semantic querying can also be used for content based integration of learning objects. Conceptual

navigation in turn means that named links can be used in navigating between learning objects. Named links

are analogous with the relationships in conceptual models.

1 INTRODUCTION

During the past few years the term learning object is

widely used in the discussion concerning

educational information systems. Generally the term

is regarded as any entity, digital or non-digital, that

may be used for learning.

Educational metadata describes any features of a

learning object. Well-designed and sufficient

metadata facilitates the learners in retrieving

relevant learning objects and aids the educational

institutions to provide suitable information about

their instruction supply. Learning object metadata is

also needed for supporting the management of

collections of learning objects, and for supporting

the decision process of the learners in looking

educational resources.

To standardize learning object metadata specific

standards are developed. LOM (Learning object

metadata standard) (LOM, 2002) defines the

structure of a meta-data instance for a learning

object. However, it does not specify the granularity

of a learning object. Fundamentally a learning object

could be a sentence, a paragraph, a topic, a section, a

chapter, a lesson, a course or even a video stream.

Using the LOM it is possible to specify for example

the grade level of a course, typical learning time of a

course, the prerequisites of a course and the

relationships of learning objects.

Dublin Core (Dublin, 2002) is a widely known

metadata standard. Its metadata elements represent

syntactical metadata, i.e., they do not describe the

content of the target. Dublin Core also includes

metadata attributes that can be used in specifying the

relationship between resources. Thorough these

attributes it is possible to define for example that a

lecture is a part of a course (IsPartOf), a course is a

version of another course ((IsVer-sionOf), a

laboratory work requires certain software

(IsRequiredBy), and a course is based on another

course (IsBasedOn).

Both LOM and Dublin Core are metadata

standards. By following these standards one can

state for example that the course Introduction to

programming precedes the course Java-

programming and that the Java laboratory work is a

part of the course Java-programming. The problem

however, is that though they allow the specification

of the relationships between the instances of learning

objects they provide no means for modelling such

relationships. As a result, we cannot for example

express semantic queries like “Give me all courses

that precede the course Java-programming” or

“Give me all the components of the course Java-

programming”. Such queries require an ontology

that gives the semantics for the learning objects

metadata expressions.

Learning object content models in turn are

developed to increase the reusability of learning

objects. They are typically taxonomies, which

365

Puustjärvi J. (2006).

THE ROLE OF LEARNING OBJECT ONTOLOGIES.

In Proceedings of WEBIST 2006 - Second International Conference on Web Information Systems and Technologies - Society, e-Business and

e-Government / e-Learning, pages 365-370

DOI: 10.5220/0001239303650370

 SciTePress

identify the components of learning objects.

However, a lack of learning object content models is

that they do not provide means for expressing the

semantics of the components of learning objects.

For example, using a learning object content

model we cannot specify whether a component, say

a course, deals with history or mathematics. And

therefore they do not allow semantic querying over

the components (e.g., querying the courses that deals

with discrete mathematics) and conceptual

navigation between the components. This is

regrettable since semantic querying and conceptual

navigation between learning objects would

significantly ease the access of learning objects.

In this article we present what kind of ontologies

are required for semantic querying and conceptual

navigation between learning objects. Essentially

semantic querying differs from traditional keyword

based searching in that searching expressions are

based on content ontologies, i.e., on the concepts of

the domain that the learning deals with. Semantic

querying is also useful tool in composing learning

objects based on their content. Conceptual

navigation in turn means that named links can be

used in navigating between learning objects. Named

links are analogous with the relationships in

conceptual scheme of databases.

The rest of the paper is organized as follows.

First, in Section 2, we give an overview of learning

object metadata standards and learning object

content models. We also illustrate the possibilities

these approaches give for expressing the

relationships of learning object instances. Then, in

Section 3, we motivate our approach by giving an

example of semantic querying and conceptual

navigation. After this, in Section 4, we show what

kinds of ontologies are required for semantic

querying and conceptual navigation. In particular,

three ontologies are presented: a content ontology,

an education ontology and an instance ontology. The

specification of these ontologies by XML-based

languages is considered in Section 5. Finally,

Section 6 concludes the paper by discussing the

advantages and limitations of our proposed

approach.

2 METADATA STANDARDS AND

LEARNING OBJECT CONTENT

MODELS

2.1 Metadata Standards

The notion of metadata (Najjar et al., 2003) has

variable interpretations depending upon the

circumstances in which it is used. Fundamentally,

metadata is data about data. It describes certain

important characteristics of its target. Equally

metadata can be described by meta-metadata, which

is descriptive information of the metadata itself. The

typical types of metadata that can be attached to

documents include document’s author, publisher,

publication date, language and keywords.

There are many organizations which standardize

metadata. The idea behind standardization is to

achieve interoperability between systems from

different origins. An important point in

standardization is that it does not impose a particular

implementation but rather a common specification

which establishes an opportunity for collaboration

by diverse groups.

Next we will shortly consider three well known

standardization efforts; Dublin Core, IMS and LOM.

Dublin Core (Dublin, 2002) is a widely known

metadata standard that has been developed since

1995. The metadata elements of the Dublin Core

represent syntactical meta-data, i.e., they do not

describe the content of the target. Originally, they

are intended to facilitate the discovery of electronic

resources from the Web. It includes 15 metadata

elements that describe the content, the intellectual

property rights and the instantiation of the object.

For example, the standard includes the following

elements: Creator, Date, Description, Subject, and

Language. Even though, the Dublin Core does not

include educational metadata elements, it has been

used as basis for many educational metadata

projects. On the other hand, proposals to extend the

standard by educational elements (e.g., Audience,

Interactivity type, and Interactivity level) have been

done.

Dublin Core also includes metadata attributes that

can be used in specifying the relationship between

resources. Thorough these attributes it is possible to

define for example that a lecture is a part of a course

(IsPartOf), a course is a version of another course

((IsVer-sionOf), a laboratory work requires certain

software (IsRequiredBy), and a course is based on

another course (IsBasedOn).

IMS (Instructional Management System Project)

(IMS, 2002) is a consortium of several educational

institutions, commercial entities, government

agencies, and developers in the area of educational

information systems. Its main aim is to develop and

promote open specifications for facilitating online

distributed learning activities such as tracking

learner progress, reporting learner performance, and

exchanging student records between administrative

systems.

IMS has been a significant contributor to the

LOM. For example, it has introduced the use of

XML for representing metadata. On the other hand,

WEBIST 2006 - E-LEARNING

366

IMS uses the LOM as its basis for metadata

specifications. For example, IMS has contributed to

LOM by introducing best practice guides for

metadata developers and implementers.

LOM (Learning object metadata standard) (LOM,

2002) defines the structure of a metadata instance

for a learning object. A learning object is regarded

as any entity, digital or non-digital, that may be used

for learning. In addition, the standard facilitates the

sharing and exchange of learning objects by

enabling the development of catalogues and

inventories while taking into account the diversity of

cultural contexts in which the learning object will be

exploited. The goals of the LOM are to enable the

learners to search and use learning objects and

enable computer agents to automatically compose

learning objects to individual learners.

Using the LOM it is possible to specify for

example the teaching or interaction style of a course,

the grade level of a course, the difficulty of a course,

typical learning time of a course, the prerequisites of

a course and the relationships of learning objects’

instances.

2.2 Learning Object Content Models

Learning Object Metadata standard does not specify

the granularity of a learning object. Fundamentally a

learning object could be a sentence, a paragraph, a

topic, a section, a chapter, a lesson, a course or even

a video stream. A small and self-contained learning

object has a good chance of reusability. On the other

hand, the complexity of composing learning objects

increases as the amount of learning objects

increases.

Learning object content models are developed to

increase the reusability of learning objects. They are

typically taxonomies, which identify the syntactical

components of learning objects. Taxonomy is a

hierarchical structure (a tree) where the relationship

between a parent and its children has some

relationship (e.g., is part of).

The SCORM (SCORM, 2005) Content

Aggregation Model is a taxonomy which is

comprised of the following levels (Figure 1): Assets,

Sharable Content Objects (SCO) and Content

Aggregations. For example, text, images, audio and

other data that can be presented in the web client are

Assets. A Sharable Content Object is a collection of

one or more assets. In order to increase the

reusability of Sharable Content Objects should be

independent of its learning context. So it can be

reused in different learning experiences to fulfil

different learning objectives. A Content Aggregation

is a structure that can be used to aggregate learning

resources in an integrated unit such as course or

chapter.

Figure 1. The structure of the SCORM Content Aggregation Model.

Content Aggregations

Sharable Content Objects Sharable Content Objects

Assets

Assets Assets

Figure 1: The Structure of the SCORM Content

Aggregation Model

3 SEMANTIC QUERYING AND

CONCEPTUAL NAVIGATION

Semantic querying and conceptual navigation allow

easy searching facilities of learning objects. Further

semantic querying can be used for automatic

composition of learning objects.

In order to illustrate semantic querying and

conceptual navigation let us consider the following

example. Assume that a learner wants to renew his

or her programming skills. Now, in order to find

appropriate course the learner performs the

following action.

First the learner asks the educational system to

display a content taxonomy Programming. Then the

system displays the taxonomy presented in Figure 2.

Then the learner chooses from the taxonomy the

concepts that should be included to the course, say

Object oriented programming, Java-programming

and C++ programming and returns them to the

system. The system then returns a course, say Object

oriented programming languages and provides links

for preliminary courses, lectures, exercises and

exercise solutions.

Programming

Database

programming

Object oriented

programming

Java-

programming

C++

programming

SQL-

programming

WWW-

programming

Figure 2. Taxonomy Programming.

Figure 2: Taxonomy Programming.

THE ROLE OF LEARNING OBJECT ONTOLOGIES

367

4 LEARNING OBJECT

ONTOLOGIES

4.1 The Goal of Ontologies

An ontology is a general vocabulary of a certain

domain (Davies et al., 2002), and it can be defined

as “an explicit specification of a conceptualization”

(Gruber, 1993). Essentially the used ontology must

be shared and consensual terminology as it is used

for in-formation sharing and exchange.

Ontology tries to capture the meaning of a

particular subject domain that corresponds to what a

human being knows about that domain (Daconta et

al., 2002). It also tries to characterize that meaning

in terms of concepts and their relationships.

Ontology is typically represented as classes,

properties attributes and values. So they also provide

a systematic way to standardize the used metadata

items.

Ontology languages provide representational

entities without stating what should be rep-resented,

i.e., they do not commit to any particular domain

(Antoniou and Harmelen, 2004). For example the

ER-model (Ullman &Widow, 1998), RDFS (RDFS,

2005), ODL (Ullman &Widow, 1998), UML,

DAML+OIL (DAML, 2005) and OWL (OWL,

2005), which define concepts such as entities or

objects, attributes and relations, are ontology

languages.

A salient feature of ontologies is that depending

on the generality level of conceptualization, different

types of ontologies are needed. Each type of

ontology has a specific role in information sharing

and exchange.

4.2 Content Ontologies

The purpose of the content ontology is to describe

the concepts of the domain in which learning take

place. So, the content ontology may for example

describe the concepts related to mathematics, history

or to computer science. To illustrate content

ontologies a simple content ontology Programming

is presented in Figure 3.

Programming

Database

programming

Object oriented

programming

Java-

programming

C++

programming

OO-

programming

SQL-

programminf

WWW-

programming

isPartOf

isSynonym

Figure 3. Content ontology Programming.

4.3 Education Ontology

Education ontology captures the entities that are

related to learning. In addition it captures the

relationships of the entities. It has a similar function

as database scheme defined by data definition

languages (e.g., ODL (Ullman & Widom, 1998)). A

difference, however, is that education ontology is

presented by ontology languages and thus it provides

syntactically and semantically richer means than

database definition languages.

In Figure 4, an Education ontology is presented.

It captures entities that are related to learning in

universities. It is presented in a graphical form but it

can also be presented in OWL. In the figure, a

relationship (property in OWL terminology) related

to the object (Class in OWL terminology) “course”

is “precedes”. In OWL one can specify that this

property is transitive. So, for example if Course A

precedes Course B and B precedes Course C, then

the system can infer that also Course A precedes

Course C. This is one feature that can be defined in

ontology languages but not in the data definition

languages developed for databases.

Figure 3: Content Ontology Programming.

WEBIST 2006 - E-LEARNING

368

lecture

course

exercise

solution

paragraph

chapter

course

book

programme

teacher

professor assistant

university

department

laboratory

course

material

animation

isPartOf

isPartOfisPartOf

isPartOf

isTypeOf

isPartOf

lectures supervises

isSuclassOf isSuclassOf

isAssociated

belongsTo

isAssociated

Figure 4. An Education ontology.

precedes

It captures entities that are related to learning in

universities. It is presented in a graphical form but it

can also be presented in OWL. In the figure, a

relationship (property in OWL terminology) related

to the object (Class in OWL terminology) “course”

is “precedes”. In OWL one can specify that this

property is transitive. So, for example if Course A

precedes Course B and B precedes Course C, then

the system can infer that also Course A precedes

Course C. This is one feature that can be defined in

ontology languages but not in the data definition

languages developed for databases.

4.4 Learning Instance Ontologies

Learning instance ontology describes the instances

of learning objects. In our graphical representation

(Figure 5) the types of each instance are also

described in the ontology. In addition the subject of

each instance can also be defined. Type is defined

by connecting an instance to the learning object

ontology (typeOf edges) while subject is defined by

connecting the instance to the content ontology

(subjectOf edges).

Learning instance ontology deviates from

content ontology and learning entity ontologies in

that it is described by the content creator (e.g. a

teacher). In other words a content creator annotates a

learning instance according to the concepts

presented in content and education ontology.

Figure 5. An example of an instance ontology.

Java-course

Course

Intoroduction

to Java

Course book

Introduction

to programming

Department of

Computer Science

Department

isAssociated

typeOf

precedes

belongsTobelongsTo

Programming

OO-

programming

Java-

programming

subjectOf

Each instance in the instance ontology in Figure 5 is

presented by an oval, and corresponding objects

(classes in OWL) are presented by a rectangle, and

they are connected by the typeOf-edge. The content

of an instance is represented by edge subjectOf edge

to an item of the content ontology which is presented

inside of circle.

The instance ontology of Figure 5 describes that

Java programming and Introduction to

programming are instances of the object (class in

OWL) Course and both courses belong to the

Department of Computer Science, which is an

instance of the object Department. The “precedes”

edge (property in OWL) between the courses Java-

programming and Introduction to programming

indicates that before the execution of the course

Java-programming the course Introduction to

programming must be executed. The ontology also

indicates that the course book Introduction to Java is

associated to the course Java-programming.

5 XML-BASED LANGUAGES

AND LEARNING ONTOLOGIES

We now give a short introduction to the XML-based

languages that we are using in specifying learning

object ontologies.

RDF provides a means for attaching semantics

(e.g., metadata values) to objects (e.g., to Java-

course). So it nicely adapts for specifying the edges

in ontologies. For example, the description (see

Figure 5) “The subject of Java-course is Java-

programming” can be expressed in a RDF-statement.

The relationship of XML and RDF is that XML

Figure 4: An Education Ontology.

Figure 5: An example of an instance ontology.

THE ROLE OF LEARNING OBJECT ONTOLOGIES

369

provides a way to express RDF-statements. In other

words, RDF is an application ox XML.

Fundamentally, RDF defines a language for

describing relationships among resources in terms of

named properties and values. It however, provides

no mechanisms for describing these properties, nor

does it provide any mechanisms for describing the

relationship between these properties and other

resources. That is the role of RDF vocabulary

description language RDF schema (RDFS, 2005). It

defines classes and properties that may be used to

describe classes, properties and other resources.

Hence, there is a straight correspondence between

RDF schema and object oriented design.

OWL Web Ontology Language (OWL, 2005) has

more facilities for expressing meaning and

semantics than XML, RDF and RDF Schema, and

thus OWL goes beyond these languages in its ability

to represent machine interpretable content of the

ontology. In particular, it adds more semantics for

describing properties and classes, for example

relations between classes, cardinality of

relationships, and equality of classes and instances.

For example, the graphical representation “Object

oriented programming is a synonym for OO-

programming” in Figure 3 can be expressed in

OWL.

6 CONCLUSIONS

Educational information systems should be designed

in a way that they provide easy access to learning

objects. Well-designed and sufficient metadata

facilitates the learners in retrieving relevant learning

objects and aids the educational institutions to

provide suitable information about their instruction

supply

To standardize learning object metadata specific

standards, such as the LOM, are developed. The lack

of the LOM, as with all metadata models, is that it

does not provide semantics for the metadata items

As a result, many useful learning object retrieval

methods such as semantic querying and conceptual

navigation cannot be implemented in educational

information systems that are based on metadata

standards.

By introducing sharable learning object

ontologies we can specify the semantics of the

metadata items. In order to give a semantics for

metadata items, we have represented a simple

ontology, called educational ontology. Further in

order that we can specify the content (subject) of an

educational material (e.g., a course) we have given

an example of a content ontology. We have also

introduced instance ontologies. Through an instance

ontology we can tie learning instances to the objects

of the education ontology and to a content ontology.

The main gain of our proposed ontologies is that

they provide a conceptual model on which semantic

querying and conceptual navigation can take place.

Semantic querying can also be used for content

based integration of learning objects. So they can

also be used for extending the function of learning

object content models which compose learning

object components solely based on their structure

without considering their content.

A drawback of our approach is that it burdens the

content creator (e.g., a teacher) in that he or she has

to annotate learning material according to the

ontologies. However, it is turned out that computer

support can alleviate this function in many ways.

REFERENCES

Antoniou, G. & Harmelen, F. (2004). A semantic web

primer. The MIT Press.

Daconta, M., Obrst, L. & Smith, K. (2003). The semantic

web. Indianapolis: John Wiley & Sons.

DAML+OIL (2005). http://www.daml.org/2001/03/daml

+oil/index.html

Davies, J. Fensel, D. & Harmelen, F. (2002), Towards the

semantic web: ontology driven knowledge

management. West Sussex: John Wiley & Sons.

Dublin Core (2002). http://www.dublin.core.org.

Gruber, Thomas R. (1993). Toward principles for the

design of ontologies used for knowledge sharing.

Padua workshop on Formal Ontology.

IMS Global Learning Consortium. (2002).

http://www.imsproject.org/metadata/index.html

LOM – Learning Object Metadata. (2002).

http://ltsc.ieee.org/doc/wg12/LOM_1484_12_1_v1_Fi

nal_Draft.pdf.

Najjar, J. Stefaan Ternier, S. & Duval, E. (2003). The

Actual Use of Metadata in ARIADNE: an

Empirical Analysis, 3rd Annual Ariadne

Conference, Belgium.

OWL –WEB Ontology Language (2005).

http://www.w3.org/TR/owl-ref/

RDFS (2005). http://www.w3.org/TR/2000/CR-rdf-

schema-20000327/

SCORM (2005). http://www.rhassociates.com/scorm.htm

Ullman, J. & Widom J. (1998) Principles of Database

Systems. Prent

ice Hall. 1998.

WEBIST 2006 - E-LEARNING

370