Research On Concept Semantic Similarity Computation In Ontology Mapping

Today, Semantic Web is a good way to solve the problem that the current Web can't process the massive information automatically. Ontology is the base of the Semantic Web because it is a good method of conceptualization of the domain knowledge. The distribution of the Web makes that different users develop their own ontologies according to their requirement. The content of these ontologies is overlapped or related in their semantic, but the description language and the model are different, So the ontology heterogeneous has generated. Ontology mapping mechanism can solved the problem well. And in the process of ontology mapping, determining the similarity of the concepts across different ontologies is the key.The Triple Matching Distance Model(MD3) is a typical method to determine the semantic similarity of concepts from different ontologies. The model is on the basis of ontology's representations, to determining the semantic similarity from three facets: lexicon matching, feature matching, and semantic neighborhood matching(semantic relation matching). The global similarity is then a weighted sum of the semantic similarity of each component. MD3 model is a good method, but it has some shortcomings.This paper proposed a synthesized approach of similarity computation aiming at the current problems in the process of ontology similarity computation. It used the Hirst-St-Onge semantic relativity algorithms to determine whether there is correlation among different domain ontologies at first. It will take account of the name, attributes, instances, relations of concepts and compare in many ways if it relevant. On the basis of analysis to the MD3 model, taking into account the influence of non_hiberarchy relations among concepts and the instances of concepts, the paper has proposed the EMD3 model(Extension of Triple Mapping Distance model). On the basis of the EMD3 model, we have proposed the ontology mapping mechanism, designed the flow of ontology mapping, and implemented the arithmetic of ontology mapping, and then have built the experiment flat of ontology mapping. Tree groups experiment results show that EMD3 model is better in the recall and precision than the MD3 model in the same condition. With the development of the ontology, it is true that the advantages of the EMD3 model will be found in the near future.