Research On Effective And Adaptive Ontology Matching Algorithm

Ontology is an explicit formal specification of a shared conceptualization. Ontologyis used to formally specify domain knowledge. Thereby it can overcome semanticheterogeneity of information exchange effectively between distributed applications.However, the subjectivity and autonomy in ontology building and using cause manyheterogeneous ontologies in various domains which has become one of the biggestbottlenecks for ontology application. Ontology matching is the most effective method forsolving this problem.Ontology matching seeks to find semantic correspondences of different ontologies.In recent years, researchers at home and abroad have gained successes by stages onontology matching. But due to the difficult matching task, there are still new challenges.This paper aims at shortages of existing matching algorithms and proposes improvementsabout three problems.（1）Proposing a method of aggregating various strategies adaptively. According tocharacteristics of different ontologies and difference of similarity matrixes, this paperutilizes harmony, richness and difference to measure credibility of each strategy so as toselect credible matchers adaptively and determine the weights of the matchers. Besides,matchers are optionally combined at the semantic level and then correct the similarity ofhigh credible mappings.（2）Proposing an efficient ontology mathcing algorithm based on graph structure.Similarity propagation condition is extended to RDF triples. Only credible similar seedsare able to propagate similarity. During iterating similarities, merely adjust thesimilarities of similar seeds and probable similar pairs. Similarity is calculated byelements’ features. This algorithm reduces computational complexity greatly. And takingelements’ semantic characteristics into account enables similarity more accurate. （3）Proposing the candidate mappings debugging and extracting technique.Candidate mappings are used to integrate ontologies as bridge rules. Diagnoseunintended entailments in the integrating ontology caused by error mappings. Findconflict mappings which cause unintended entailments. Then based on conflict mappings,improve naive descending algorithm which is used to extract mappings so that it canextract consistent mappings.Finally, based on the above work, the paper adopts Java and uses Jena API todesign the algorithms. And integrate the algorithms for the ontology matching system.OAEI data set and evaluation methodology are used in the experiment. Experimentsresult shows that the algorithms are effective and improve the precision whilemaintaining recall, thus get a better mapping result.