THE ORM MODEL AS A KNOWLEDGE REPRESENTATION

FOR E-TUTORIAL SYSTEMS

Thanaporn Leelawatananon

Faculty of Information Technology,

King Mongkut’s Institute of Technology, Ladkrabang, Bangkok 10520, Thailand,

Suphamit Chittayasothorn

Faculty of Engineering,

King Mongkut’s Institute of Technology, Ladkrabang, Bangkok 10520, Thailand,

Keywords: Object Role Model, Knowledge-based System, Knowledge Representation, e-Tutorial.

Abstract: At present information technology plays important roles in teaching and learning activities. E-learning

systems have the potential to reduce operating costs and train more people. Teachers and students do not

have to be in the same place at the same time and the students have the opportunity to perform self-studies

and self-evaluation using e-tutorial systems. E-learning systems could be considered expert systems in the

sense that they provide expert advice in particular subjects of studies to students. The exploitation of

knowledge base and knowledge representation techniques is therefore vital to the development of e-learning

systems. This paper presents the development of a knowledge-based e-tutorial system that uses the Object

Role Model (ORM) as its knowledge representation. The system provides Physics tutorials. It was

implemented in Prolog and the knowledge base is on a relational database server.

1 INTRODUCTION

Nowadays, the most common underlying data

management facilities of e-learning systems are still

the conventional files and record structures. Most e-

learning systems interact directly with files and

tables without knowledge representations. Recently,

the use of ontology and knowledge representation

models in e-learning systems appears in literatures.

Resource Definition Framework (RDF) ) (Brickley

and Guha, 2000) and Ontology Interchange

Language (OIL) (Studer, 2000) are used to

represent knowledge in e-learning systems based on

the Semantic Web (Stojanovic et al., 2001).

This paper presents an e-tutorial system which

uses the Object Role Model (ORM) (Halpin, 2003)

as a knowledge representation. The ORM model is a

well-established conceptual schema model

originated in Europe during the 1970s and was

originally called NIAM (Verheijen and van Bekkum,

1982). It is a popular conceptual schema model for

relational database design. The transformation from

an ORM conceptual schema diagram to relational

database schemas guarantees the project/join normal

form (5NF).

Conventional knowledge-based systems that

implement the knowledge base on a relational

database use relational database schemas as

predicate structures structures (Norrie et al., 1995;

Albernethy and Altman, 1998; Maier et al., 2002). A

tuple is perceived as a predicate instance. The

problems of this approach are that a tuple of a

relation may contain several facts only one of which

is relevant and a tuple may have null values. Using

the ORM conceptual schema as a knowledge

representation solves these problems. An ORM fact

type is a predicate structure and a fact instance

becomes a predicate instance. Fact instances always

have the truth value “true” following the 2-value

logic and null value is not permitted in ORM. These

properties make ORM a suitable knowledge

representation for knowledge base systems

implemented on large shared relational databases.

479

Leelawatananon T. and Chittayasothorn S. (2004).

THE ORM MODEL AS A KNOWLEDGE REPRESENTATION FOR E-TUTORIAL SYSTEMS.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 479-484

DOI: 10.5220/0002655504790484

 SciTePress

2 OBJECT ROLE MODEL

ORM is a fact-based conceptual schema model. Its

main constructs are entity types, label types, fact

types, reference types, subtype hierarchies and some

static integrity constraints. It demonstrates clear

distinction between the concept of entity types and

the naming of the entity types (label types).

An entity type is an object of interest. An entity

type may have several label types (value types)

associate with it via reference types. Label types are

naming of entity types and there is no concept of

attributes in ORM. A one-to-one reference type is

chosen to be the unique identifier of the entity type.

Other reference types remain alternative identifiers

of the entity type. Figure 1a shows an entity type

STUDENT together with its label types SNAME

and ID#. Figure 1b shows that ID# is chosen to be

the unique identifier of STUDENT. The unique

identifier is shown in brackets inside of the entity

type. SNAME remains in a reference type.

Relationships between entity types are called fact

types. These fact types must be elementary (cannot

be further decomposed). N-ary fact types are also

allowed. A uniqueness constraint enforces

uniqueness of entity instances participated in the fact

type. A transformation algorithm from an ORM

allowed. A uniqueness constraint enforces

uniqueness of entity instances participated in the fact

type. A transformation algorithm from an ORM

conceptual schema to relational database schemas

takes advantage of this non-decomposable property.

Each fact type is perceived as a relational database

schema in 5NF. Fact types associated with a

common entity type and have the uniqueness

constraint enforced on the entity type form a

relational schema in 5NF (Halpin, 2003). Figure 1c

shows ORM fact types and corresponding relational

database schemas.

3 ORM AS A KNOWLEDGE

REPRESENTATION MODEL

From the previous section, the ORM model is

perceived as a relational database design tool that

guarantees the 5NF relational database schemas. In

this project, we use it as the logical structure that our

e-learning system works on. ORM is the knowledge

representation of our system.

Figure 1: a) An entity type and label types. b) An entity type together with its unique identifier and a

reference type. c) An ORM conceptual schema diagram and corresponding relation schemas.

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

480

ELECTRIC_FIELD

CHARGE_DISTRIBUTION

(CDISTRIBUTION)

DISTRIBUTION_SYMBOL

(SSYMBOL)

CONDUCTOR

(CID#)

TOTALCHARGE_FORMULAR

(TCFORMULAR)

GAUSSIAN_SHAPE

(GSHAPE)

SURFACE_SIDE

(SSIDE)

SURFACE_AREA

(SID#)

VARIABLE

(VCODE#)

VARIABLE_CONSTAN T

(VCONSTANT)

VARIABLE_DIRECTION

(VDIRECTION)

VARIABLE_UNIT

(VUNIT)

MULTIPLE_VECTOR_SYMBOL

(MVSYMBOL)

MULTIPLE_VECTOR_TYPE

(MVTYPE)

MULTIPLE_VECTOR_FO RMULAR

(MVFORMULAR)

VARIABLE_SYMBO L

(VSYMBOL)

Figure 2: ORM diagram for the Gaussian electric field calculation

THE ORM MODEL AS A KNOWLEDGE REPRESENTATION FOR E-TUTORIAL SYSTEMS

481

This approach has the advantage over the use of

conventional knowledge representations such as

Frames and Semantic Networks (Waterman, 1999)

when implemented in relational database environ-

ments. There are no transformation algorithms that

guarantee minimum redundancies and the 5NF

(project/join normal form) from Frames and

Semantic Networks. This is due to the fact that they

are pointer-based representations. Most artificial

intelligence systems implement them as Lisp

programs. ORM, on the other hand, was developed

for database modeling. The implementation of an

ORM knowledge base using commercially available

relational DBMS enables the knowledge base to be

shared by many user applications. such as recovery

control, concurrency control, indexing and query

optimizations are readily available for our

knowledge-based e-tutorial system. Figure 2 shows

the ORM diagram for the Gaussian electric field

calculation topics of the e-tutorial system. Another

important point of using the ORM model as a

knowledge representation is that each predicate

instance corresponds with an ORM fact instance, not

the entire tuple of a relation.

This is Figure 2:

ORM diagram for the Gaussian electric field

calculation very useful since each tuple could

contain many facts. Classical and more recent

interfaces between expert systems and database

systems refer to a relation as a predicate instance

(Wang, 2000; Nick et al., 2001). This is not realistic

in practice because there could be irrelevant facts on

each tuple. It is proposed that a predicate instance

refers to a fact instance of an ORM fact type.

4 ORM META CONCEPTUAL

SCHEMA

An ORM meta conceptual schema is a conceptual

schema that describes the ORM conceptual schema

model. Since the users’ ORM schema for domain

knowledge must be stored on the database, a set of

system tables is required to keep the information

about the users’ conceptual schemas. The meta

conceptual schema is transformed into relational

schemas for the system tables that keep information

about the users’ ORM schemas. Figure 3 shows the

ORM meta conceptual schema which is used by our

e-tutorial system.

5 AN E-TUTORIAL SYSTEM FOR

PHYSICS

The prototype e-tutorial system presented in this

paper gives Physics tutorials. It assists student’s

work on Physics exercise questions and evaluates

students’ understanding of the topic. The exercises

are grouped in chapters. For each exercise, the

system asks questions to guide the student to the

solution of the problem. The questions are

sequenced in the following order: questions on the

formulae used for the given exercise, questions

about relevant variables, questions about the main

knowledge of the exercise and the application of

formulae to obtain the result.

During a working session the system analyses

the answer to each question to evaluate the student’s

understanding of the topic and shows the marks and

evaluation result to the student. The system interacts

with its ORM knowledge base to obtain related

knowledge for the guidance and evaluations.

The system is implemented in WinProlog

(Steel, 2000) and the underling DBMS is MS

SQL*Server (Rebecca, 2000). ORM fact types are

implemented as views and Prolog retrieves the

content of the views when it consults the database.

This means that the Prolog program is not aware of

the underlying tables and refers to fact type views

only. Each Prolog predicate instance an ORM fact

instance, not the entire tuple of a relation. Figure 4

shows data flow diagrams that describe the e-

tutorial system. Figure 5 shows a sample study

session in Gaussian electric field calculation. The

feature of our prototype e-tutorial system is

comparable to other systems such as ANDES system

(Joel, 2000), and another web-based tutorial system

for engineering, mathematics and science subjects

(Scott and Stone, 2000). Our prototype system

guides the students step by step while some of these

systems such as ANDES give the full guidance first

and then let the students solve the problem

afterwards. However, the features of the e tutorial

system is not the main issue here. It is the ability to

store Physics knowledge on relational database

using the ORM model as the knowledge

representation.

6 CONCLUSION

The e-tutorial system presented in this paper uses the

ORM conceptual schema model as its knowledge

representation. The system refers to fact instances of

fact types when it analyzes student’s answers and

evaluates the level of understanding of the topic.

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

482

This is a new approach in knowledge-based system

design and implementation on large shared

databases. Developers use SQL to access the

knowledge base. So, this approach is easy to

Figure 3: An ORM Meta Conceptual Schema

Rol e

(Nr)

Pl ays

Ent i ty Type

(Name)

Label Type

(Name)

Spans

has

Constrai nt

(Nr)

KOC

(Code)

has

Constrai ntType

(Code)

NOR

(Name)

{Unary,Binar y,N_ary}

{UI,UE,MR}

has

AOR

(Code)

{ MN, N M, MM, N N}

UniqueIdenti find

(Name)

has

Relati onshi p

(Nr)

KOC=Kind of Cadinality

NOR=Name of Relationship

AOR=Amount of Role

UI=Uniqueness Internal

UE=Uniqueness External

MR=Mandatory Constraint

MN=Many-to-Many

NM=One-to-Many

NN=One-to-One

MN=Many-to-One

Figure 4: A data flow diagram for the e

tutorial system

THE ORM MODEL AS A KNOWLEDGE REPRESENTATION FOR E-TUTORIAL SYSTEMS

483

implement from the developer’s point of view. The

knowledge can be shared, reuse, and manage in

multi-user environment by database servers.

REFERENCES

Abernethy, N., Altman, R., 1998. Sophia : Providing

Basic Knowledge Service with A Common DBMS,

Proceedings of the 5th KRDB Workshop.

Brickley, D. and Guha, R., 2000. Resource Description

Framework (RDF) Scheme Specification 1.0.

Candidate recommendation: World Wide Web

Consortium. See http://www.w3.org/TR/2000/CR-rdf-

scheme-20000327.

Halpin, T.A., 2003. Information Modeling and Relational

Databases Using ORM with ER and UML: Morgan

Kaufmann.

Horrocks, I., Fensel, D., Broekstra, J., Decker, S.,

Erdmann, M., Goble, C., van Harmelen, F., Klein, M.,

Staab, S., and Maier, F., Nute, D., Potter, W., Wang,

J., Twery, M., Rauscher, M., Knopp, P., Thomasma,

S., Dass, M. and Uchiyama, H., 2002.

PROLOG/RDBMS Integration in the NED Intelligent

Information System. In the proceedings of the Tenth

International Conference on Cooperative Information

Systems, Irvine, California.

Joel, A., 2000. A Computer Tutorial System for

Introductory Physics Courses, See:

http://www.physics.rutgers.edu/~shapiro/RutgK15/Rut

gK15.pdf

Nick B., Ioannis V., Ahmed K. and Elias N., 2001,

InterBase: Integrating a knowledge Base System with

a Multidatabase System for Data Warehousing, See:

http://lpis.csd.auth.gr/publications/interbase-kb.pdf

Norrie, M., Reimer, U., Lippuner, P., Rys, M., Schek, H.,

1995. Frames, Objects and Relations : Three Semantic

Levels for Knowledge Base Systems. Swiss Priority

Programme in Computer Science No. 5003-34347.

Rebecca Riordan, 2000. Microsoft® SQL Server™ 2000

Programming Step by Step.

Scott, N., and Stone, B., 2000. A Flexible Web-based

Tutorial system for Engineering, Maths and Science

Subjects, Global J. of Engng. Educ., Vol.2,No.1,

Australia.

Steel, D., 2000 , Win-PROLOG Win 32 Programming

Guide, See : http://www.lpa.co.uk

Stojanovic, L., Staab, S., Studer R., 2001. eLearning based

on the Semantic Web. In: Webwet 2001-World

Conference on the WWW and Internet, October 23-27,

Orlando, Florida,US.

Studer, R., 2000. OIL: The Ontology Inference Layer.

Technical report: University of Manchester / Vrije

Universiteit Amsterdam. See

http://www.ontoknowledge.org/oil/.

Verheijen, G. M. A., and J. van Bekkum, 1982. NIAM: An

Information Analysis Method. Information Systems

Design Methodologies: A Comparative Model. Proc.

IFIP WG 8.1 Working Conference: North-Holland.

Wang L. and Wang M., 2000, Relational knowledge

Representation Model in Intelligent Tutor System,

See:

http://www.pku.edu.cn/academic/xb/2000/_00e512.ht

Waterman, D. A., 1999. A Guide to Expert Systems:

Addison-Wesley.

Figure 5: A sample study session in Gaussian electric field calculation

ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS

484