An adaptive ontology matching approach

Ontology matching is a promising way to harmonize heterogeneous ontologies in various applications. In this paper, we propose an adaptive ontology matching approach to match different kinds of ontologies automatically according to their structural features. Various similarity measures are designed to calculate similarities from different ontology features. If there is abundant structural information in ontologies, it combines different similarities in a parallel manner and aggregates them with a differentor-based similarity aggregation strategy that assigns weights for each row in a similarity matrix. If the ontologies are composed of concept hierarchies, a semantic inductive similarity flooding algorithm is utilized to match ontologies in graph. For extracting alignment, a greedy extraction algorithm is proposed that considers the maximum values in rows or columns in similarity matrix. Finally, we test the proposed approach on OAEI test for evaluation. The experimental results demonstrate that differentor-based aggregation strategy outperforms other existing aggregation strategies. The proposed alignment extraction algorithm performs better than the naïve descending extraction algorithm. In the most recent OAEI2011, the proposed approach outperforms all other systems on the new dataset benchmarks2.