Natural Communication Between Person and Computer
Bertram C. Bruce Bolt Beranek and Newman Inc. Cambridge, MA
This chapter discusses a computer natural-language system called "HWIM," which accepts either typed or spoken inputs and produces both typed and spoken rxesponses. There are three major purposes of the chapter: (1) to present HWIM as an example of a relatively complete language system in which one can see how the many components of language processing interact; (2) to describe technical problems that arose during HWIM's development and that are illustrative of major concerns of Artificial-Intelligence research; and (3) to discuss the structure of HWIM as a model for discourse processing.
This chapter discusses some general issues of language comprehension and production through examination of a complete natural-language system called HWIM (for "Hear what I mean"). HWIM was developed over a 5-year period as the BBN speech-understanding system. It was designed to understand natural language (typed or spoken); to answer questions, perform calculations, and maintain a data base; and to respond in natural language (typed and spoken). Both its inputs and outputs used a relatively rich grammar with a 1000-word vocabulary. Utterances were assumed to be part of an on-going dialogue so that HWIM had to have a model of the discourse and the user.
HWIM was designed to serve as an assistant for the manager of a travel budget. The task of the system was to assist the travel-budget manager, helping to record the trips taken or proposed and to produce such summary information as