Fuzzy set and semantic similarity in ontology alignment

Abstract

A challenge for the Semantic Web is enabling information interoperability between related but heterogeneous ontologies. Ontology alignment (OA) addresses this challenge by identifying correspondences between entities in different ontologies. The traditional OA evaluation strategy uses a gold standard reference alignment created by a domain expert. The problem is that often a reference alignment may not exist. The current use of semantic similarity measures in the OA process and proposals for their use in the OA evaluation task are presented. Many semantic similarity measures are derivative of fuzzy set similarity measures. A general semantic alignment quality (SAQ) measure is developed and used on the alignment results of 10 different OA systems produced on the anatomy track of the 2010 OA evaluation initiative. The SAQ results indicate much variation in performance depending on the selected semantic similarity measure. Problems with using several semantic similarity measures in SAQ are further investigated and the findings are discussed.