Using Domain Ontologies to Discover Direct and Indirect Matches for Schema Elements

Automating schema matching is challenging. Previous approaches (e.g. [DDH01, RB01]) to automating schema matching focus on computing direct matches between two schemas. Schemas, however, rarely match directly. Thus, to complete the task of schema matching, we must also compute indirect matches. In this paper, we focus on recognizing expected values as a technique to discover many direct and indirect matches between a source schema and a target schema. This technique relies on domain ontologies, which must be handcrafted. The benefits appear to justify the cost as demonstrated in the experiments we have conducted over a real-world application. The experiments show that this technique increases the results by an increase about 20 percentage points, yielding an overall result above 90% precision and recall for both direct and indirect matches.