EUROSPEECH 2001 Scandinavia
7th European Conference on Speech Communication and Technology
2nd INTERSPEECH Event

Aalborg, Denmark
September 3-7, 2001

                 

A Data Selection Strategy for Utterance Verification in Continuous Speech Recognition

Hui Jiang, Frank K. Soong, Chin-Hui Lee

Bell Labs, Lucent Technologies Inc., USA

In this paper, we propose the concept of rival for verifying hypothesis in speech recognition. A likelihood ratio test, based on the rivals model, are investigated for utterance verification in continuous speech recognition. We present an efficient method to train rival model automatically from training data as well as a single pass strategy of utterance verification, namely verification-in-search, for continuous speech recognition. Some preliminary experiments on DARPA Communicator travel task have shown the rival models give better verification performance in terms of identifying mis-recognized words from the output of our baseline recognizer.

Full Paper

Bibliographic reference.  Jiang, Hui / Soong, Frank K. / Lee, Chin-Hui (2001): "A data selection strategy for utterance verification in continuous speech recognition", In EUROSPEECH-2001, 2573-2576.