Scalable matching of industry models - A case study

Abstract

A recent approach to the problem of ontology matching has been to convert the problem of ontology matching to information retrieval. We explore the utility of this approach in matching model elements of real UML, ER, EMF and XML-Schema models, where the semantics of the models are less precisely defined. We validate this approach with domain experts for industry models drawn from very different domains (healthcare, insurance, and banking). We also observe that in the field, manually constructed mappings for such large industry models are prone to serious errors. We describe a novel tool we developed to detect suspicious mappings to quickly isolate these errors.