KID Designing a Knowledge-Based Natural Language Interface

Abstract

T his article presents a knowledge-based approach to designing portable natural language interfaces to database systems-an approach called KID (knowledge-based interface to database systems). To free users from needing to know about query languages and databases, and to provide users with access to databases in natural language, KID incorporates a knowledge base called the world model containing a domain model, linguistic knowledge, and database mapping knowledge. The world model is easy for users to define and understand; to transport KID to a new domain, the user has only to build a new world model for the domain. Using the world model, KID processes user queries transparently enough to enable easy user debugging of the world model. Since we have encapsulated domainspecific knowledge into the world model, KID is portable. This article will discuss world model design, and will examine KID's knowledge-based processing of queries using the world model. Numerous natural language interfaces to database systems have been developed in an attempt to make such systems more user friendly.1l2 Natural language interfaces must incorporate a knowledge base containing knowledge about a domain of discourse, databases, natural language, and a query language-thereby freeing users from needing to know about query languages and databases and enabling users to access databases in natural language. Moreover, applying natural language interfaces to new domains requires construction of new knowledge bases for those domains. Recent work focuses on portable natural language interface systems.3'4 However, no work has focused on the issues involved when users design knowledge bases by themselves to transport natural language interface systems to new domains-although one exception discusses some of these issues.5 -Gio Wiederhold Associate Editor

A knowledge engineer based on natural language understanding for an expert system: DOMES

“…As reported by Ledgard et al (1980), an interface system with natural language understanding will increase efficiency and reduce errors noticeably. There has been a considerable amount of effort devoted in developing an interface system with natural language understanding among the artificial intelligence applications community (Davis 1979;Peterson 1980;Good et al, 1984;Wilensky et al, 1984;Ishikawa et al, 1987).…”