An Improved Discrete Particle Swarm Optimization For Ontology Matching

Semantic web can make the information on the web understood by the computer. This makesthe automation processing, sharing and reuse of information come true. The definition of ontologyin computer science is“an explicit specification of a conceptualization”. Ontology is the base of theSemantic Web. All of these reasons result a mass disappear of heterogeneous ontologies: the targetsof constructing ontology are various; the constructors of the ontology may have differentknowledge background; the methods of constructing ontology are different from each other andthere is no a uniform standard for ontology construction. There are many different ontologies in thesame or similar domain and this results the problem of ontology heterogeneity. Ontologyheterogeneity deviates from the original target of an explicit specification of a conceptualization. Inorder to solve the problem of ontology heterogeneity, we need to find the semantic relationsbetween the ontologies and make an ontology matching.Firstly, this paper puts forward a frame of ontology matching based on improved discreteparticle swarm optimization. The entities either appear, either do not arise in the mapping results, sothe problem of ontology matching is discrete. This paper properly developed the particle swarmalgorithm model for the purpose of using the particle swarm algorithm on the progress of ontologymatching. At the same time, this paper takes the problem of ontology matching as an optimizationproblem for the purpose of using the particle swarm algorithm on the progress of ontology matching.The target of optimization is: (i) identify a set which contains the correct matching results, and (ii)make the number of the correspondences in the set as max as possible. This paper learns from theparticle warm algorithm which used by Correa et al. in the data mining domain. This paper changesthe rule that each particle has a fixed length and redesigns the strategy of particle’s iterative andupdating progress. In the algorithm designed by this paper, each particle represents a candidate setof matching results between the original ontology and the target ontology. Each particle has a speedwhich contains the likelihoods. The convergence of the swarm is guided by proportional likelihoodvalues assigned to each correspondence, based on the best alignments that the swarm and eachparticle individually have so far discovered.Furthermore, this paper gives a realization of designing basic matching methods. This paperdesigns five basic matching algorithms: (i) the basic matching algorithm based on linguistictechnology, (ii) the basic matching algorithm based on WordNet, (iii) the basic matching algorithmbased on information retrieval, (iv) the basic matching algorithm based on comparing ontologystructure, and (v) the basic matching algorithm based on comparing the relations between theconcept and property. More specifically, the basic matching algorithm based on linguistictechnology contains (i) string directly comparing, (ii) substring comparing, and (iii) calculation of Levenshtein distance. And then, this paper design a compose strategy for the purpose of composingthe results calculated by the basic matching method to get the entry’s adaptive value which neededby the particle swarm algorithm.