Fuzzy Protg for Fuzzy Ontology Models

Hanne GHORBEL --- Afef BAHRI --- Rafik BOUAZIZ MIRACL Laboratory, Faculty of economics and management of Sfax - Tunisia
E-mail

: [email protected] ;

[email protected] ; [email protected]

Abstract. Fuzzy Protg is a semi-automatic collaborative tool for the construction of fuzzy ontology models, built as a Protg 3.3.1 tab plug-in. Fuzzy Protg is an extension of the well known ontology editor Protg for which we have defined new meta-classes to allow the definition of parameterized membership functions. Fuzzy Protg also gives support to instantiate fuzzy concepts and roles and allows automatic computing of membership degrees. Finally, Fuzzy Protg allows querying fuzzy ontologies based on fuzzy criteria. We present in this paper the internal architecture of Fuzzy protg and we give some details on its implementation and the way we use it to build and validate fuzzy ontologies. Key words. Semantic Web, Fuzzy Ontologies, Fuzzy Protg, Inference in Fuzzy Ontology, Interrogation for Fuzzy Ontologies.

1 Introduction
The web content is, for the most part, subject to imperfection (i.e. imprecision, uncertainty, ambiguity). Semantic Web ontologies are based on crisp logic and do not provide welldefined means for expressing imperfection. How to represent non crisp data, such as an expensive book, a good customer within the ontology definition. The idea of fuzzy set and fuzzy logic theory was first proposed by Zadeh [Zadeh, 75], as a mean of handling uncertainty. The appearance of a wide variety of methodologies for the construction of fuzzy membership functions motivates the application of fuzzy set theory in a great number of application fields. In this paper, we talk about the integration of fuzzy logic in ontology in order to define which we call fuzzy ontology. Ontology tools are most based on crisp logic and do not provide well-defined means for expression fuzzyness. We propose in this paper a framework for fuzzy ontology building, Fuzzy Protg, as an extension of the wellknown ontology editor Protg. Then we describe how Fuzzy Protg has been designed an implemented. The paper is structured as follows. Section 2 describes the internal architecture of the framework and the implementation details of the building modules, showing how to use Fuzzy Protg to define fuzzy ontology components. Section 3 defines a rule inference engine implemented with Jess. Section 4 presents a flexible query tool. Finally, in section 5, we conclude and present some topics for further research.

2 Architecture of Fuzzy Protg

Fuzzy ontology has been introduced by Straccia [06] to represent knowledge in all domains in which the concepts to be represented have imprecise definitions. Fuzzy concepts and roles are considered as fuzzy sets. Thus an instance does not fully belong or not to a given fuzzy concept (resp. fuzzy role) but has a membership degree being an instance of that concept (resp. role). For instance, the fuzzy concept Young_Person is defined as follow: Young_Person = Person Age.young The linguistic term Young may be defined by a trapezoidal function as shown graphically in Figure 1. Mathematically, it is defined as follows:
Young

Young(Age) =

(Age 12) / (18 12) if 12 <= Age <= 18 1 if 18 < Age <= 30 (45 - Age) / (45 30) if 30 < Age <= 45 0 if Age < 12 Or Age > 45

Age 12 18 30 45 Fig.1. Example for trapezoidal membersh ip function

The principle idea of our framework is the use of parameterized membership functions to allow automatic attribution of membership degree. Then we will be able to perform inconsistency checking as membership degrees are not randomly added as some works do. We define a new metaclass named Fuzzy_Class in the Protg metaclasses hierarchy (basic class is the class Thing). The figure 2 shows the metaclasses hierarchy of Fuzzy Protg. To use these metaclasses, we have to import the specific meta-data using Meta data plug-in. So, the Fuzzy Meta-Classes become classes systems, users of the framework can neither modify nor remove them (cf. Figure 3). Indeed, users may define a given fuzzy ontological component as a subclass of one of the Fuzzy classes defined in the model (i.e. Fuzzy_Class_Ti, Fuzzy_Class_Trz, Fuzzy_Class_L, Fuzzy_Class_R).
: Thing

: System

: Meta-Class

: Constraints

: Annotation

: Relation

: Fuzzy-Class

: Class

: Slot

: Facet

: Directed-Binary-

: Fuzzy-Class-Trz

: Fuzzy-Class-Tri

: Fuzzy-Class-L

: Fuzzy-Class-R

: Fuzzy-Relation

: Fuzzy-Relation-Trz

: Fuzzy-Relation-Tri

: Fuzzy-Relation-L

: Fuzzy-Relation-R

Fig.2. Fuzzy Protg Hierarchy metaclasses

Fig.3. Meta-classes of Fuzzy Protg

3 Inference and validation of fuzzy instances

Instantiation is considered as a type of inference. In the case of fuzzy ontologies, instantiation consists on determining the degrees of membership of instances to fuzzy concepts or roles. We define in Fuzzy Protg a rule based inference engine for automatic membership degree computing. The module is implemented under the Jess language and we use the Jess Tab for compilation. We give in the following, the definition of rules used for membership degree computing of a concept defined with trapezoidal membership function.

Rule 1: Determinate of degree of membership if a <= X <= b

(if (and(slot-get <--Instance--> X >= Fuzzy_a) (slot-get <--instance--> X <= Fuzzy_b)) then (slot-set <--Instance--> Fuzzy:Degree_Of_Membership (/ (- (slot-get <--Instance--> X) Fuzzy_b)) (- (Fuzzy_b Fuzzy_a))) (if (and(slot-get <-Instance--> X >= Fuzzy_b) (slot-get <--instance--> X <= Fuzzy_c)) then (slot-set <--Instance--> Fuzzy:Degree Of Membership (- (slot-get <--Instance--> X) Fuzzy b)))

Rule 2: Determinate of degree of membership if b < X <= c

(if (and(slot-get <--Instance--> X > Fuzzy_b) (slot-get <--instance--> X<= Fuzzy_c)) then (slot-set <--Instance--> Fuzzy:Degree_Of_Membership 1.0)

Rule 3: Determinate of degree of membership if c < X <= d

(if (and(slot-get <--Instance--> X > Fuzzy_c) (slot-get <--instance--> X <= Fuzzy_d)) then (slot-set <--Instance--> Fuzzy:Degree_Of_Membership (/ (+ ( - slot-get <--Instance--> X) Fuzzy_d)) (- (Fuzzy_d Fuzzy_c)))

Rule 4: Determinate of degree of membership if X > d And X < a

(if (and(slot-get <--Instance--> X <= Fuzzy_a) (slot-get <--instance--> X >= Fuzzy_d)) then (slot-set <--Instance--> Fuzzy:Degree_Of_Membership 0.0))

4 Fuzzy ontology querying

Fuzzy Protg contains equally a module for fuzzy ontology querying. When a user asks a query about the instances of a given fuzzy concept, we have to return those instances that have a membership degree to the corresponding concept great or equal to a given threshold. In order to do that, we propose to realize a pretreatment on the instances of fuzzy concepts before query processing. We use Jess project to classify fuzzy instances in order to eliminate the instances that have a membership degree less or equal to a certain threshold. We can then query fuzzy ontologies using Queries plug-in of Protg. We give in the following, respectively, the pretreatment (i.e. the classification) of fuzzy instances using Jess Project, and an example of a query definition with Queries plug-in of Protg.
Individuals classification of a fuzzy ontology - Jess Project Defclass <Name-New-Set> (is-a <Name-Fuzzy-Concept>) If (slot-get <--Instance--> Degree-Of-Membership >= s) then (make-instance <Instance> of: <Name-New-Set>) With s is a threshold of membership Treatment the user request by using Queries plug-in

Degree_Of_Membership

0.8

Choose New Set

Choose Slot

Constraints specification

Figure.4. Fuzzy ontology classification and querying

Figure 4 presents then in the part "Treatment the user request by using Queries plug-in" an example of the requests which can be treated by Fuzzy Protg. In this example, the user needs to find the list of the "Young Person". With this intention, he must choose the fuzzy subset (i.e. fuzzy concept) Young_Person in the part Class. Thereafter, he must also state that its search criteria are the degree of membership. He has then to put in the part Slot the property "Degree_Of_Membership". Finally, he must specify the minimal degree of membership to seek (in our example 0.8). With all these stages, we succeeded to treat a great range of fuzzy requests users by using the degrees of membership already calculated with the module inference and validation.

5 Conclusion and further work

In this paper, we have presented the internal architecture of Fuzzy Protg and the way we can use it to implement fuzzy ontologies. Fuzzy Protg aimes to be a framework which ensures the operationalization of fuzzy ontologies (building, instantiation, validating, inconsistency checking and querying). The actual version of Fuzzy Protg hasbeen used for the implementation of fuzzy ontologies of Employees competences management [Ghorbel, 08b]. In the future works, we project to enrich the querying module and complement the inference engine to allow inconsistency checking of fuzzy ontologies.

6 References
[Ghorbel, 08a] H. Ghorbel, A. Bahri and R. Bouaziz ; Les langages de description des ontologies : RDF & OWL ; Proc. of Gnie lectrique et informatique GEI2008 ; pages 597-606 ; Sousse, Tunisie ; Mars 2008. [Ghorbel, 08b] H. Ghorbel, A. Bahri and R. Bouaziz ; A Framework for Fuzzy Ontology Models ; Proc. of journes Francophones sur les Ontologies JFO2008 ; Lyon, France ; pages 21-30 ; Dcembre 2008. [Stoilos, 06] G. Stoilos, U. Straccia, G. Stamou et J. Pan, General concept inclusions in Fuzzy Description Logic , Proc. of the 17th European Conference on Artificial Intelligence (ECAI-06), 2006. [Straccia, 06] U. Straccia, A Fuzzy Description Logic for the semantic Web , Proc. in Capturing Intelligence: Fuzzy logic and the semantic Web, Elie Sanchez ed., Elsevier, 2006. [Zadeh, 75] L. Zadeh, The concept of a linguistic variable and its application to approximate reasonning ; Proc. of Information Science ; Volume 4, Issue 4 ; pages 301-357 ; 1975.