Natural language access to multiple databases: A model and a prototype

A model is proposed that defines "context" for natural language queries in a multiple database environment and a suggested implementation for this model is described. The approach developed in this study relaxes the assumption usually made for natural language query systems that the context, in the form of a specific database, is given. The formal model defines a universe of discourse consisting of multiple contexts, each of which is represented by a database, which consists of a dictionary, a thesaurus, relations, columns, and values. A mechanism for identifying and selecting the appropriate context for a query and its implementation as a prototype are described. The integration of this mechanism into the general model of natural language query processing is indicated and implementational issues are discussed.