Pattern-based core word recognition to support ontology matching

Ontology matching is a crucial issue in semantic web and data interoperability. In this paper, we describe a core word based method for measuring similarity from the semantic level of ontology entities. In ontology, most of labels of entities are compound words rather than single meaningful words. However, the main meaning usually is represented by one word, which is called core word. The core word is learned by investigating certain patterns, which are defined based on part of speech (POS) and linguistics knowledge. Also, the other information is noted as complementary information. An algorithm is given to measure the similarity between a pair of compound words or short texts. In order to support diverse situation, especially when no core words could be recognized, non semantic based ontology matching techniques are applied from lexical and structural aspects of ontology. The described method is tested on real ontology and benchmarking data sets. It showed good matching ability and obtained promising results.