Ontology alignment using Rough Sets

Abstract

Ontology alignment has been one of the most significant research issues to bridge the semantic gap between heterogeneous systems toward Semantic Web. Many proposed systems in recent years, use multiple similarity measures and variety of aggregating algorithms to map entities from two different ontologies. However, the real issue arises when alignment systems turn out to be uncertain about entities which are neither completely similar nor dissimilar. This paper presents an ontology alignment system which uses a combination of various similarity measures in order to map entities from two different ontologies. Rough Sets have been used to deal with uncertainties during the mapping process by defining the similarity measures as the attributes of entities. The proposed alignment system employs the combination of string based, linguistic based and structural matchers. The structural matcher compares the super-classes, sub-classes and properties of two class entities. We use WordNet to compare the linguistic similarity between entities in the mapping process. The performance of proposed system is evaluated in terms of precision and recall and compared with existing state of the art alignment systems. For this purpose, the Ontology Alignment Evaluation Initiative (OAEI) benchmark group-3 test ontologies have been used. Experimental tests show that the performance our proposed alignment system is comparable with existing alignment systems.