Matching of Ontologies with XML Schemas Using a Generic Metamodel Christoph Quix, David Kensche, and Xiang Li RWTH Aachen University, Informatik 5 (Information Systems), 52056 Aachen, Germany {quix,kensche,lixiang}@cs.rwth-aachen.de Abstract. Schema matching is the task of automatically computing correspon- dences between schema elements. A multitude of schema matching approaches exists for various scenarios using syntactic, semantic, or instance information. The schema matching problem is aggravated by the fact that models to be matched are often represented in different modeling languages, e.g. OWL, XML Schema, or SQL DDL. Consequently, besides being able to match models in the same meta- model, a schema matching tool must be able to compute reasonable results when matching models in heterogeneous modeling languages. Therefore, we developed a matching component as a part of our model management system GeRoMeSuite which is based on our generic metamodel GeRoMe. As GeRoMe provides a uni- fied representation of models, the matcher is able to match models represented in different languages with each other. In this paper, we will show in particular the results for matching XML Schemas with OWL ontologies as it is often required for the semantic annotation of existing XML data sources. GeRoMeSuite allows for flexible configuration of the matching system; vari- ous matching algorithms for element and structure level matching are provided and can be combined freely using different ways of aggregation and filtering in order to define new matching strategies. This makes the matcher highly con- figurable and extensible. We evaluated our system with several pairs of XML Schemas and OWL ontologies and compared the performance with results from other systems. The results are considerably better which shows that a matching system based on a generic metamodel is favorable for heterogeneous matching tasks. 1 Introduction Integration of information systems is a major challenge that has been addressed in sev- eral disciplines such as database and semantic web research. One of the key issues in integration is creating a mapping between the data models of the systems involved. This work is, for example, required if the data from different data sources must be merged in a data warehouse or if two e-business systems must communicate with each other. Schema matching is the task of identifying a set of correspondences (also called a morphism or a mapping) between schema elements. Many aspects have to be consid- ered during the process of matching, such as data values, element names, constraint information, structure information, domain knowledge, cardinality relationships, and so on. All this information is useful in understanding the semantics of a schema, but it R. Meersman and Z. Tari et al. (Eds.): OTM 2007, Part I, LNCS 4803, pp. 1081–1098, 2007. c  Springer-Verlag Berlin Heidelberg 2007