On the Use of Syntax and Semantics in a Speech Understanding System.

摘要

The document represents an attempt to produce a system,called the HEARSAY system,capable of accepting speech input,understanding spoken messages,and generating voice responses. This system has as subparts a speech recognition system, a question-answering system,and a speech generation system. A task is a system capable of answering questions, obeying commands,and/or accepting facts for future use. In a speech understanding system, a task has a corresponding task language,which is a formal language with an associated semantic interpretation. Two tasks are implemented in the HEARSAY system. The first,voice-chess,accepts spoken chess moves and speaks response moves. The second,voice-programming,accepts simple spoken algebraic expressions and speaks their values. Some preliminary results obtained on the HEARSAY system are presented. (Modified author abstract)