Abstract
The first step in interoperating among multidatabases is semantic integration: Producing attribute correspondences that describe relationships between attributes or classes in different database schemas. Dynamic integration requires the ability to automatically extract database semantics, express them as metadata, and match semantically equivalent data elements to produce attribute correspondences. This process cannot be pre-programmed since the information to be accessed is heterogeneous. An architecture supporting dynamic integration is presented. Semint, a tool for automated semantic integration, that helps database administrators generate attribute correspondences, is discussed. A novel framework for dynamic integration and a query language for multidatabase systems that uses Semint as part of a complete semantic integration service are introduced. The framework supports dynamic integration as well as incremental integration. The advantages of the framework in an environment where full integration is not desired or complete knowledge of the databases to be integrated is unavailable are shown.