Reliability-based similarity aggregation in ontology matching

“Ontology matching” aims to find semantic correspondences between entities of different ontologies. Achieving semantic interoperability in building the Semantic Web highly relies on matching ontologies. It has been shown that integrating multiple individual matchers to explore different aspects of similarities between ontologies leads to better results than just using one matcher. Thus, similarity aggregation is an important and yet difficult step in developing ontology matching systems. Previously, a harmony based adaptive aggregation approach (called HADAPT) proposed aggregation of different similarity measures using an adaptive method based on their harmonies to estimate the performance of similarities. In this paper, we first discuss a shortcoming of the HADAPT and then introduce a reliability-based similarity aggregating method that can fix the problem. Our method extracts important information from the similarity matrices, which were ignored by previously proposed similarity aggregation methods. Experimental results show that our aggregation method outperforms other existing aggregation methods, including HADAPT, on OAEI benchmark tests.