Reconfiguration of speech recognizers through layered-grammar structure to provide ease of navigation and recognition accuracy in speech-web.

摘要

Developing speech interfaces to large knowledge bases is a new and challenging problem. There is a need for a solution to provide access to large knowledge bases and high recognition accuracy. A partial solution to this problem is to distribute the knowledge base into a network of speech-accessible units of knowledge. But as the number of such units increases the recognition accuracy decreases and navigation among these units becomes difficult. In this thesis, a new technique is investigated. The new technique is based on a layered grammar structure and modification of the unitu27s input language to provide high recognition accuracy with ease of navigation among units. This technique is a step towards a solution for high recognition accuracy and distribution transparency with ease of navigation for large knowledge bases. A prototype has been implemented to demonstrate the efficiency of the layered grammar based approach. Paper copy at Leddy Library: Theses u26 Major Papers - Basement, West Bldg. / Call Number: Thesis2000 .Q74. Source: Masters Abstracts International, Volume: 40-03, page: 0727. Adviser: Richard Frost. Thesis (M.Sc.)--University of Windsor (Canada), 2001.