A Harmony based Adaptive Ontology Mapping Approach
Ming Mao, Yefei Peng, Michael B. Spring · 2008
Abstract- Ontology mapping seeks to find semantic correspondences between similar elements of different ontologies. Ontology mapping is critical to achieve semantic interoperability in the WWW. Nowadays most ontology mapping approaches integrate multiple individual matchers to explore both linguistic and structure similarity of different ontologies. Thus how to effectively aggregating different similarities is pervasive in ontology mapping. In current aggregation methods, people either have to manually set parameters in aggregation function or need "ground truth " in advance for machine learning based parameter optimization. Both of them have limitation. In this paper, we propose a measure harmony, which is the normalized number of mapping pair that suggests an unambiguous one-to-one mapping, and a harmony based adaptive ontology mapping approach, which can automatically adjust parameters of three kinds of similarities (i.e., edit distance based similarity, profile similarity and structure similarity) in aggregation functions according to different mapping tasks without given any ground truth. Experimental results show the harmony is indeed a good estimator of the performance (i.e., f-measure) of different similarities, and the harmony based adaptive aggregation method outperforms all other existing aggregation methods on OAEI benchmark tests.