Retrieving spoken documents by combining multiple index sources

Abstract

This paper presents domain-independent methods of spoken document retrieval. Both a continuous-speech large vocabulary recognition system, and a phone-lattice word spotter,