Spoken variable initiative dialog: an adaptable natural-language interface

Abstract

Several obstacles have prevented spoken natural-language systems from providing the required performance, including inflexibility, ineffective goal-seeking, and poor speech recognition. However, these problems can be circumvented by embedding the speech recognition technology within a dialog processing mechanism. Such a variable initiative dialog is a major advance that lets natural-language systems communicate effectively with novices and experts. Based on a theory of natural-language dialog that addresses these issues, the author has implemented a system that uses spoken natural language to help users repair electronic circuits. The integrated dialog-processing model combines a domain problem solver, a general subdialog mechanism, and knowledge about the user to provide timely and coherent assistance to the user. The robust parsing and language-understanding mechanism helps the system to correctly determine the meaning of utterances in spite of misrecognitions. Results indicate that commercial application of this technology is within reach, and should stimulate thought about how to further improve the quality of spoken natural-language interaction. >