Research On Ontology Matching Algorithm Based On Markov Network

Ontology is a formal, explicit specification of a shared conceptualization. At present,ontology has been applied in such fields including semantic web, artificial intelligence,knowledge engineering. Ontology is the core component of semantic web, however, thesubjectivity and autonomy in ontology building and ontology using cause the phenomenonof ontology heterogeneity. Ontology matching is the most effective solution to it, whichestablishes the interoperable interaction between the application systems.Ontology matching is uncertain and the logic relationships between the mapping pairsare complex. The probabilistic graphical models are effective tools to solve the problem ofuncertainty inference and data analysis. The features of markov network enable it to bettersupport the non-causal dependency relationship between the mapping pairs. Therefore, anontology matching algorithm based on markov network is proposed, which focuses on theuncertainty of ontology matching. The main contents are listed as follows:(1) One model for ontology matching based on markov network is proposed. Thesimilarity matrix is computed using several traditional algorithms, then the anchor nodesare found. The structures of all the cliques and the corresponding potential functions aredefined using similarity propagation and structure consistency constraint. The markovnetwork is constructed according to the similarity matrix, the anchor nodes and the cliques.Finally, the mapping results are obtained by doing approximate reasoning.(2) One improved loopy belief propagation algorithm is proposed for ontologymatching. There are some circles in markov network for ontology matching. Loopy beliefpropagation to address ontology matching can not guarantee that getting convergenceperformance of algorithm and obtaining good approximate results. Improve the posterioriprobability expression, optimize message-passing schedule especially for ontologymatching model, use the cutset-subgraph method to increase performance of algorithm and to obtain better matching results.(3) Design and implement an ontology matching system based on markov network.Analyse the system features and the general framework, design and implement all themodules of the system.This paper uses data sets provided by OAEI and experiments on the algorithms. Theexperimental results prove that the algorithms are effective and they improve the precisionand recall rate, thus get a better quality of ontology matching.