Optimizing Alignment Selection in Ontology Matching via Homomorphism Constraint

Abstract

Ontology matching is an important issue for integrating information from distributed ontologies on the Web. While lots of research is related to similarity measures, little attention has been paid to the methods for selecting alignments from a similarity matrix. In this paper, we propose two alignment selection methods based on homomorphism constraint and weak constraint on homomorphism respectively. Experiments on various OAEI tests show that the two methods have an advantage when the matching ontologies have sufficient subsumption relations while performing competitively in other cases. Finally, we design a strategy to dynamically choose a suitable method according to the characteristics of the compared ontologies. Experimental results demonstrate that this strategy leads to a stable advantage in both precision and F1-measure in average.