Ontology matching based on Probabilistic Description Logic

A lot of attention has been devoted to probabilistic methods for discovering semantic mappings between ontologies. Despite impressive theory foundation, these methods usually require massive data instances to learn the mappings, which are not always available in practice. In this paper we present the Probabilistic Description Logic based Ontology Matcher (PDLOM) for discovering such mappings using the inference service provided by Probabilistic Description Logic (P-CLASSIC), which allows for computing the probability of a concept description in an ontology. Unlike other probability based approaches, PDLOM only needs a probability distribution over primitive concepts instead of massive data instances of all concepts in the ontologies. We propose to exploit a search engine to get the probability distribution required by Probabilistic Description Logic. We evaluate our algorithm and compare against the schema matching tool COMA++. PDLOM shows an average improvement of 6% in quality over COMA++. Key-Words: Ontology matching, Probabilistic Description Logic, Bayesian networks, Search engine, Page count