ONTOLOGY MATCHING BASED ON BIPARTITE GRAPH

Ontology matching is to find out similar concept-pairs from hetero-geneous ontologies, and to create the maximum amount of bijections, so-called 1-1 mappings for tkem. Recently many efforts on ontology matching have been carried out, with some problems such as efficiency and completeness. To this end, this paper proposes a method of Ontology Matching based on Bipartite Graph. This method firstly calculates the similarity of two concepts from different ontolo-gies by comparing their instance and property sets. Then, a process of mapping discovery is carried out in order to create mappings for the concept-pairs, in which the similarity of concepts is more than or equal to the given threshold provided by ontology expert. To solve the problem of mapping a concept to mul-tiple concepts, This method generates bipartite graphs based on the mapping result, in each of which all the concepts are interconnected with a non-1-1 map-ping. Finally, this method applies an improved Hungarian Algorithm to search the maximum matchings of the bipartite graphs, in order to accomplish the task of ontology matching. Experimental results indicate that: (1) this method has a good performance on the public dataset (Benchmark) provided by Ontology Alignment Evaluation Initiative; (2) it works well when matching the ontologies coming from the real world.