The Research Of Multi-Strategies Based Ontology Matching

The Semantic Web is the expansion of existing Web, its goal is to enable the information on the Web have semantics which computer can understand. But ontology describes these semantics, and it plays the vital role in the Semantic Web. As a result of the Semantic Web distributed characteristic, a large number of data are coming from different ontologies inevitably, but in the integration of heterogeneous ontologies, finding the matching between ontologies' terms is needed. There are already a lot of different matching strategies, but to users, it is difficult to determine which strategy is most effective.In this paper, the ontology matching strategies recommending system is given. First of all, a few ontology fragment pairs are selected from original ontology pairs, and given semi-automatically or manually the matching relationships. Then they are matched by different matching strategies. Finally, according to estimate matching strategies'report provided a result of recommended strategies. The fragment of the original ontology is only a small part of the entire ontology. Compared to the whole ontology, the effort and complexity to add mappings have been relatively small.In this paper, learning has been adopted in the realization of ontology matching. Firstly, in the strategy of based on the concept instances ontology matching, using the joint distribution of learning technology. Secondly, in the definition of the concept of ontology ontology matching, a learning method based on the WordNet is utilized. Finally, in the ontology matching strategies recommending, through the matching assessment of the original ontology fragments, the recommended strategy is gained by learning, thus the ontology matching is completed.In the design of Ontology matching strategy recommending, a conception of degree of matching strategy adaption (MSA) is proposed. According to the requirements of the different focus on ontology matching, different MSA is generated, thus the top n strateies recommending are given.