Research On An Efficient Ontology Matching Algorithm

Ontology is an important component in semantic web. In semantic web, ontologies are decentralized in distribution. Ontologies from a certain domain suffer from disparities of terms and structures which severely hinder interactions between ontologies because they are created by semantic web users independently. Nevertheless, tasks in semantic web involve ontologies work together more or less in order to be fulfilled. Therefore, ontologies are not born to be suitable for these tasks. Ontology matching is aimed at overcoming this embarrassing situation, and so that enable interoperations and cooperation among ontologies. Ontology matching is the process of finding correspondence of entities from two ontologies by algorithm. By ontology matching, large scale tasks which need lots of ontologies cooperate with each other can be achieved successfully. Hence, researches on ontology matching are significant for semantic web.Current ontology matching approaches which utilize ontology's entity hierarchy tree firstly calculate all entities'similarity from two different ontologies, and select valid matches by setting threshold value, and then use the entity hierarchy tree to conduct matches'filtering from bottom to top or iterative matching. However, these approaches fail to use the entity hierarchy tree reasonably and therefore lead to computation redundancies and also increase the probability of generating erroneous matches.After analyzing current ontology matching approaches, this thesis proposed an efficient ontology matching method. This method first casts matching from top to bottom of the entity hierarchy tree, which makes good use of the tree structure to decrease computation redundancies and meanwhile reduces the probability of making wrong matches, and then a re-matching method is used to overcome the drawbacks of incomplete matching caused by structural difference. Finally, this thesis improves similarity computation, a core process in ontology matching, so that matching performance can be enhanced via improving similarity computation.Finally, this thesis adopts Java, uses program package Jena and Alignment API to conduct simulation experiment to verify the method proposed in this thesis. The experiment results show that the new ontology matching order can get good performance and improve matching efficiency, and the performance can be better via improving similarity computation.