The speech recognition technique has used in the various fields, with the progress of it and the needed functions are different for purposes. For example, speech dictators need to recognize whole sentences, but spoken dialogue systems need to recognize important words rather than whole sentences. In this paper, we propose a method of modifying language model for improving recognition correct rate of important words. This method multiplied probabilities of unigram and bigram of important words by constant weight and made important words to easy to be recognized. And we examine differences of recognition result in changes of a set of important words. As a result, the correct rate of important words increased when we multiplied probabilities of important words by weight.
Cite as: Shirosaki, Y., Kikuchi, H., Shirai, K. (1999) Improving recognition correct rate of important words in large vocabulary speech recognition. Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999), 511-514, doi: 10.21437/Eurospeech.1999-131
@inproceedings{shirosaki99_eurospeech,
author={Yasuo Shirosaki and Hideaki Kikuchi and Katsuhiko Shirai},
title={{Improving recognition correct rate of important words in large vocabulary speech recognition}},
year=1999,
booktitle={Proc. 6th European Conference on Speech Communication and Technology (Eurospeech 1999)},
pages={511--514},
doi={10.21437/Eurospeech.1999-131}
}