Ontology Mapping Based On Concept Lattices

With the development of Internet, information on network has been quicklyincreasing. Due to distribution of web, different organizations inevitably construct theirown ontology to describe data. This causes the existence of different ontologies even ina same field, which is called ontology isomerism. In order to reach the goal ofknowledge sharing and interaction, ontology isomerism is emergent to solve. Ontologymapping offers support for solving ontology isomerism problem.Based on deep analysis on existing traditional and FCA-related ontology mappingmethods, this paper presented a FCA-based ontology mapping infrastructure. Theoptimization of concept similarity measure and discovery of more types of mappingrelations are focused. Based on the concept lattice theory, irreducible attributes andasymmetry function are merged to design an improved concept similarity measure,which is further used to form a new infrastructure for computing individual setsimilarity. In the design of a concept lattice-based fragmental ontology mapping method,several algorithms in concept lattice theory are efficiently used to discover the mappingrelations1-1,1-n, m-n and1-null.The whole design is programmed in Java. The results of experiments indicate thatwhen the size of formal context is smaller than20×20and attributes reasonably describeobjects, the F-Measure of the improved concept similarity model is higher than theset-based method by14.7%and higher than the concept lattice-based symmetric methodby14.4%. For individual set similarity, the result of asymmetric concept lattice-basedmethod is also better than the symmetric one and Jaccard whose F-Measure are lowerby11.1%and26.4%respectively but it cost more time than Jaccard. However, ifthreshold is0.55, the concept lattice-based asymmetric method is not better than thesymmetric one. The fragmental method based on concept lattice can discover moretypes of mapping relations in a more complete way.