AN ALGORITHM FOR BUILDING INFORMATION SYSTEM’S
ONTOLOGIES
Mohamed Mhiri, Sana Chabaane, Achraf Mtibaa, Faïez Gargouri
MIRACL Laboratory, ISIM Institute, BP 1030-3018, Sfax. Tunisia
Keywords: Ontology, Information systems design, concept, semantic relationships, conceptual relationships.
Abstract: Current applications' modelling becomes increasingly complex. Indeed, it requires a hard work to study the
particular studied field in order to determine its main concepts and their relationships. The conceptual
representations (CR) results of the modelling of such applications can contain structural and semantic errors
which are not detectable by current CASE. The solution that we propose is to associate an ontology, for the
studied field, as a help to the designers during IS modelling steps. Building such information system’s
ontologies require the use of an approach allowing the determination of the concepts and the relationships
between these concepts. Using ontologies makes it possible to ensure conceptual representations' semantic
coherence for a given field. In this paper, we propose an algorithm for building an information system’s
ontology based on the comparison between the concepts and using a set of semantic relationships.
1 INTRODUCTION
Information systems are increasingly complex in
particular with the enormous growth of the volume
of data, the extraordinary evolution of the
technologies and the evolutionary requirements of
the users. Consequently, current applications
require an enormous effort of design and
development. In fact, such applications require a
detailed study of their fields in order to define their
concepts and to determine the concepts’
relationships. On the other hand, applications
become more and more cooperatives using data from
heterogeneous sources, in particular in the case of
the Web applications (e-learning, e-commerce,
semantic Web, etc). The designers are also
confronted with different conflits. A typology of
these conflicts is given in (Ouksel, 1999).
In the literature, several approaches were
proposed to solve the previously enumerated
problems in order to assist the designers in their
tasks of modelling, mainly to represent knowledge
during the process of design.
Recently, as an evolution and generalization of
those approaches, several authors proposed the
ontology concept. Ontology is now considered as the
most complete and adequate way for resolving the
major modelling conflicts. Indeed, an ontology is an
explicit specification of a conceptualization (Gruber,
1993).
In this paper, we present an algorithm for
building information system's (IS) ontologies. It is
organised as follows. In the second section, we
present our approach for building information
system's ontologies and the different semantic
relationships used. The third section presents our
framework and the corresponding algorithm for
building IS’s ontologies. We conclude this paper by
giving our future works.
2 OUR APPROACH FOR
BUILDING IS’S ONTOLOGIES
Our ontology building is based on the three steps
proposed by (Leclère, 2002) and which we adapt to
IS (Mhiri
1
, 2005). The conceptualization step
consists in identifying the knowledge contained in a
particular field. The ontologization step consists in
modelling, in a generic language, the formal
properties of the considered field. The last step, the
operationalization supports an inference mechanism
allowing the reasoning implementation.
An ontology’s modelling requires the
introduction of new semantic and conceptual
relationships which can exist between the concepts.
467
Mhiri M., Chabaane S., Mtibaa A. and Gargouri F. (2006).
AN ALGORITHM FOR BUILDING INFORMATION SYSTEM’S ONTOLOGIES.
In Proceedings of the Eighth International Conference on Enterprise Information Systems - ISAS, pages 467-470
DOI: 10.5220/0002449304670470
Copyright
c
SciTePress
A lexical study of different domains allows as to
identify the following relationships (semantic and
conceptual ones) which can exist between the
concepts. They are described as follows.
- Synonymy relationship, exists between two
different concepts which express the same meaning.
The mutual implication, i.e. the fact that one of the
terms implies the other, and that the second implies
the first, can be seen like a proof of synonymy.
- Equivalence relationship, a concept can express
in a particular context the same meaning with
another concept.
- Identity relationship, exists between two
identical concepts having the same meaning.
- Homonymy relationship, the same concept can
have two different meanings.
- Antonymy relationship, between two concepts
indicates that the negation of the first implies the
assertion of the other; we can’t deny both at the
same time.
- is_a Relationship, it is a relationship of
classification between a general concept and a
specific one.
- Classification relationship, it is a particular case
of the is_a relationship. It can be expressed by the
kind_of relationship.
- Aggregation relationship represents the
set/elements relationship.
- Composition relationship: it is a strong
aggregation in which a concept can be composed of
one or several other concepts.
In the following section, we present an algorithm
of IS's ontologies building.
3 AN ALGORITHM FOR
BUILDING IS’S ONTOLOGIES
Our proposed algorithm is based on four steps
(Figure 1).
These steps are assisted by the designers
allowing the corpus modelling, the checking and the
correction of the obtained results.
3.1 Concepts and Their
Relationships Extraction
The determination of the concepts and the
relationships between them are based on several CR
presenting a given field (e-commerce for example).
The extraction of the concepts consists in the
extraction of the classes’ names, their properties
(attributes and operations) and their relationships.
We define a set of rules, presented at the
following section, allowing the comparison between
the concepts of the different CR.
3.2 Comparison Between the
Concepts of the Different CR
The comparison step is based on the concepts’
semantic. This semantic is expressed by each
concept’s attributes and operations. The following
algorithm determines the semantic relationships
between CR concepts; it is represented by the
following rules, where we consider C1 and C2 as
concepts. We define the following functions :
Name(C) : the concept’s name
Attribute(C) : the set of all C attributes
Operation(C) : the set of all C operations.
Rule 1 :
Identity (C1, C2) ⇔
Name(C1) = Name(C2) ^
Attribute(C1) = Attribute(C2) ^
Operation(C1) = Operation(C2).
Rule 2 :
Homonymy (C1, C2) ⇔
Name(C1) = Name(C2) ^
Attribute(C1) ≠ Attribute(C2) ^
Operation(C1) ≠ Operation(C2).
Rule 3 :
Synonymy (C1, C2) ⇔
Name(C1) ≠ Name(C2) ^
Attribute(C1) = Attribute(C2) ^
Operation(C1) = Operation(C2).
Rule 4 :
Kind_Of(C1, C2) ⇔
Name(C1) ≠ Name(C2) ^
Attribute(C1) χ Attribute(C2) ^
Operation(C1) χ Operation(C2).
Rule 5 :
Equivalence(C1, C2) ⇔
Figure 1: Building steps of IS ontology.
Addition new concepts
The CR source modelling (corpus)
Extraction of the concepts and the
relationships between their concepts
Comparison between the concepts (attributes
and operations)
Ontology
RC i
ICEIS 2006 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
468
Name(C1) ≠ Name(C2) ^
Operation(C1) χ Operation(C2).
Rule 6 :
Name(C1) ≠ Name(C2) ^
Attribute(C1) ≠ Attribute(C2) ^
Operation(C1) ≠ Operation(C2)⇔
ambiguity (C1, C2).
We can propose to the designer two
solutions:
1. Add the concept introduced like
a new entry to our ontology.
2. Choose a semantic relationship
in the proposing list of the
concepts composing our ontology.
Rule 7 :
Antonymy(C1, C2) ⇔
Name(C1) ≠ Name(C2) ^
Attribute(C1)= ! Attribute(C2) ^
Operation(C1) = ! Operation(C2).
All these rules are used to elaborate the comparison
algorithm.
Comparaison between two
concepts C1, C2
Comparaison between the attributs
and the operations of C1 and C2
Comparaison between the attributs
and the operations of C1 and C2
Identity (C1, C2)
Synonymy
(C1, C2)
KindOf (C1, C2)
Name (C1) = Name (C2)
YES NO
Attrib(C1) = Attrib(C2) and Operat(C1)=Operat(C2)
Attrib(C1) included in Attrib(C2) and Operat(C1) included in Operat(C2)
Antonom y(C1, C2)
designer
propositions
Homonymy
(C1,C2)
Attrib(C1)=NOT Attrib(C2) and Operat(C1)=NOT Operat(C2)
Equivalence
(C1, C2)
Operat(C1) included in Operat(C2)
Attrib(C1) = Attrib(C2) and Operat(C1)=C2Operat(C2)
Figure 2: Comparison between the concepts of the
different CR.
In the following section, we represent our
ontology with an ontology’s definition language
(ODL).
3.3 UMLOnto Ontology
Representation
The representation of our ontology requires the use
of a comprehensible and semi-formal language. For
that, we proposed a language UMLOnto (Mhiri
2
,
2005), based on UML. UMLOnto allows
representing the concepts as well as the relationships
between these concepts.
The diagram obtained is called concepts
diagram. This diagram can be represented by an
XML file by adding new tags supporting the
characteristics of ontologies. Indeed, XML is a
standard language supported by the major current
CASE especially using UML.
We used the tag <concept > representing a
concept which can be a class or an association. A
concept is characterized by its attributes and its
operations. The relationships between concepts can
be:
-conceptual relationship <association>,
<aggregation>, < composition >, < inheritance >,
- semantic relationship which can be synonymy,
homonymy, equivalence, kind_of, antonomy.
The name of this ontology is characterized by a
tag < name_Ontology >.
We also developed a framework to implement
our approach. It is based on ArgoUML
1
, which is an
open source software using UML (figure. 3).
Figure 3: Framework interface.
In the following section, we present the fourth
step allowing the update of our ontology.
3.4 Ontology Update
The update of our ontology consists in the iterative
using of the algorithm of comparison with the CR
provided by the designer with the previous version
of ontology.
1
http://argouml.tigris.org/
AN ALGORITHM FOR BUILDING INFORMATION SYSTEM’S ONTOLOGIES
469
The ontology update must preserve its structure.
One can have two types of updates:
- The corrective update,
- The evolutionary update.
This step is still in the experimentation in our
group.
4 CONCLUSION
In this paper, we presented an algorithm for building
IS' ontologies. It allows the extraction of the
concepts and the relationships between concepts of
different CR. Then, we used an algorithm of
comparison between the concepts based on semantic
relationships. This comparison makes it possible to
determine the relationships between the concepts of
the different CR. To represent our ontology, we
proposed UMLOnto a language which is
comprehensible by the designers. This ontology is
expressed with XML by adding new tags allowing
the exchange between the CR. Our ontology
building approach is progressive and iterative.
Like prospective for this work, we will formalize
the concepts diagram with language Z and to check
the coherence of our ontology. Then, we will define
functionalities of an ontology maintenance and will
integrate them in CASE.
REFERENCES
Gruber, T.R., 1993.Toward Principles for the Design of
Ontologies, Used for Knowledge Sharing. Stanford
Knowledge Systems Laboratory.
Leclère, L., Trichet, T., Frûst, F., 2002. Construction of an
ontology related to the projective geometry. RFIA 13th
congrès des Reconnaissance des Frames et
Intelligence Artificielle, France.
Mhiri
1
, M., Mtibaa, A., Gargouri, F., 2005. Towards an
approach for building information
systems’ontologies”. FOMI’2005,1st workshop
Formal Ontologies Meet Industry, Verona, Italy, June,
9-10.
Mhiri
2
, M., Mtibaa, A., Gargouri, F., 2005. UMLOnto:
Towards a language for the specification of
information systems’ontologies. SEKE’05, the
Seventeenth International Conference on Software
Engineering and Knowledge Engineering, Taiwan,
China, July, 14-16, p.743-746.
Ouksel, A., Sheth, A., 1999. Semantic Interoperablity in
Global Information Systems. ACM SIGMOD Record,
March, Vol 28, N°1, p. 5-12.
ICEIS 2006 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION
470
