Alignement d’ontologies de domaine

Abstract

Ontology alignment is a complex task based on the definition of similarity measures. Many studies have been done. Since only one type of measurement is often insufficient to detect a similarity, different tools that combine several measurements in a sequential way have been proposed: this combination is performed a priori and can not be modified. Our goal is to design a domain ontology alignment system to achieve the goal of aligning ontologies of the same domain to allow other systems to communicate with each other using shared terms. The alignment between two ontologies (matching) consists of producing a set of correspondences between the entities. These entities can be concepts, properties or even instances. However, the automatic generation of correspondences between two ontologies is of an extreme difficulty which is due to divergences (conceptual, habits, etc.) between different communities of development of ontologies. In addition, the problem of alignment is acute when the number and volume of data patterns are important. Indeed, in real application domains where ontologies are large and complex, the requirements of the Résumé 10 execution of time and memory space are the two significant factors that directly influence the performance of an alignment algorithm. Mots clés : alignement d’ontologies, ontology mapping, hétérogénéité, médiation de context, web sémantique.