Comparing similarity combination methods for schema matching

Abstract

A recurring manual task in data integration or ontology alignment is finding mappings between complex schemas. In order to reduce the manual effort, many matching algorithms for semi-automatically computing mappings were introduced. In the last decade it turned out that a combination of matching algorithms often improves mapping quality. Many possible combination methods can be found in literature, each promising good result quality for a specific domain of schemas. We introduce the rationale of each strategy shortly. Then we evaluate the most commonly used methods on a number of mapping tasks and try to find the most robust strategy that behaves well across all given tasks.