The SRI speech understanding system
Donald E. Walker · IEEE Transactions on Acoustics Speech and Signal Processing · 1975
This paper describes the structure of the second version of the Stanford Research Institute (SRI) speech understanding system and presents the data gathered on its performance. The system is distinctive in the way that knowledge from various sources is coordinated by a "best-first" parser to predict the sequence of words in an utterance, and in the use of word functions-programs that represent the acoustic characteristics of a word-to test the predictions.