Research On Multi-strategy Ontology Matching

In the current era of knowledge economy, knowledge management has become animportant approach to improve the technology, business, strategy intelligence and the corecompetitiveness of an organization. With the development of Internet technology and theemergence of Semantic Web technology, knowledge management has entered a newdevelopment stage. As the foundation of the Semantic Web, ontology can effectivelyexpress and query the knowledge, and is the basis of realizing intelligence knowledgemanagement. But due to the distributed character and autonomy of ontology construction,and ontology evolution, there are a lot of heterogeneous ontologies in each area. Ontologymatching is an effective way to solve this problem.Scholars at home and abroad have gained achievements on ontology matchingresearch, matching strategy based on linguistic characteristics, structural characteristics,and external resources have been peoposed. But single matching strategy uses only specificinformation and has limited effect, most systems combine a variety of strategies together inorder to achieve better matching effects. The existing serial and parallel multi-strategyontology matching have problems such as repeated calculation, low matching efficiencyand the difficulty to aggregate multiple similarity matrixes. This paper designs a hybridmatching process based on the combination of ontology feature information, and improvedthe matching method based on structure calculation. The main contents are as follows:First, a multi-strategy ontology matching process based on the combination ofontology feature information is presented. The information which is available in ontologymatching process is divided into two types, elemental level and structural level. Wedetermine whether the two entities are matched according to the combination of the resultsof the preliminary strategy matchers, and dynamicly select the next matching strategy forentity pairs which can not be judged to be matched. The method determines the corresponding matching algorithm according to the difference of entities’ information,which makes the similarity calculation more specific.Secondly, this paper improved the matching method based on structure information.Tree kernel function method is introduced to calculate the similarity of the ontologystructure to inspect the reliability of the matching method based on the external structure. Acalculation method of the richness of ontology internal structure information is put forwardin order to inspect the reliability of the matching method based on the internal structure.Then adjust the weights of the two matching methods adaptively according to thecharacters of nodes information.Finally, the paper uses Java language to realize the multi-strategy ontology matchingsystem (MSOMS). OAEI data sets and evaluation methodology are used in the experimentto test the matching effect and performance of the system. Experiments prove that ouralgorithm can effectively found matching entities between heterogeneous ontologies, andhas high recall and precision.