Planning and Reformulating Queries for Semantically-Modeled Multidatabase Systems

With vast amounts of information available from various sources, integrating data from multiple databases is an important problem. The SIMS project attacks this problem using a variety of Arti cial Intelligence techniques, including planning, knowledge representation, problem reformulation, and learning. To integrate multiple databases, the user provides a semantic model of the application domain and then uses this model to describe the contents of the available databases. Given a query, the system uses a planner to decide which databases must be queried and in what order the queries should be executed. This paper focuses on the query planning problem | the selection of appropriate data sources and ordering the accesses to them, and on the reformulation of queries | the use of knowledge both about the domain and the databases to modify queries to make the retrieval plans for them more e cient.