Discovering interdatabase resemblance of classes for interoperable databases

The integration of database schemas into a federated one involves a detection phase where the similarities that exist among the classes of the different databases must be discovered in order to determine their semantic relationships. This phase is typically characterized by the complexity of the comparison task and the deep knowledge about the semantics of the databases that is required to perform it. The authors present an approach to handle this problem by upgrading the semantic level of the local schemas and guiding the search of the comparison process by the structure of the generalization/specialization semilattices and aggregation graphs of the resulting rich schemas. Unnecessary as well as most promising comparisons are identified and a systematic procedure to analyze the rich component schemas in a meaningful way is given. A reduction of the complexity is achieved. The methodology serves as the basis for a tool that semiautomatizes the process. The expressiveness of the canonical model plays a central role.