Ontology mapping discovery with uncertainty
in Fourth International Conference on the Semantic Web (ISWC-2005, page 537--547. (2005)Abstract
Resolving semantic heterogeneity among information sources is a central problem in information interoperation, information integration, and information sharing among websites. Ontologies express the semantics of the terminology used in these websites. Semantic heterogeneity can be resolved by mapping ontologies from diverse sources. Mapping large ontologies manually is almost impossible and results in a number of errors of omission and commission. Therefore, automated ontology mapping algorithms are a must. However, most existing ontology mapping tools do not provide exact mappings. Rather, there is usually some degree of uncertainty. We describe a framework to improve existing ontology mappings using a Bayesian Network. Omen, an Ontology Mapping ENhancer uses a set of meta-rules that capture the influence of the ontology structure and the semantics of ontology relations and matches nodes that are neighbors of already matched nodes in the two ontologies. We have implemented a protype ontology matcher using probabilistic methods that can enhance existing matches between ontology concepts. Experiments demonstrate that Omen successfully identifies and enhances ontology mappings significantly.