Motivation and overview of SPEECHLIS: An experimental prototype for speech understanding research

SPEECHLIS is a research prototype of an intelligent speech understanding system which makes use of advanced techniques of artificial intelligence, natural language processing, and acoustical and phonological analysis in an integrated way to determine the interpretation of continuous speech utterances. This paper describes a number of the characteristics of the speech understanding task which influence the ways in which syntactic, semantic, pragmatic and lexical knowledge interact with acoustical and phonological information in the process of understanding speech utterances. The focus is on what the different knowledge sources have to contribute at different points in the analysis and the organization of a computer system to combine these different sources of information into an integrated system.