Problems in Natural Language Communication with Computers

This paper gives an overview of the problems involved in the construction of a computer-based question-answering system designed to interact with the user in English. The system is viewed as containing five distinct parts: a parser, a semantic interpreter, an information storer, an information retriever, and an English output generator. There is a need for extensive interaction among these subsytems, and between the subsytems and the user. Examples are given of the type of processing done by each subsystem, and the nature of the possible interactions. The syntactic analysis described is based on a Chomsky type of transformational grammar. The semantic store is characterized by a form of the predicate calculus, with additional algorithms for computation, and structures designed for fast access to relevant data.