In this paper we define a number of confidence measures derived from an acceptor HMM and evaluate their performance for the task of utterance verification using the North American Business News (NAB) and Broadcast News (BN) corpora. Results are presented for decodings made at both the word and phone level which show the relative profitability of rejection provided by the diverse set of confidence measures. The results indicate that language model dependent confidence measures have reduced performance on BN data relative to that for the more grammatically constrained NAB data. An explanation linking the observations that rejection is more profitable for noisy acoustics, for a reduced vocabulary and at the phone level is also given.
Cite as: Williams, G., Renals, S. (1998) Confidence measures derived from an acceptor HMM. Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998), paper 0644, doi: 10.21437/ICSLP.1998-434
@inproceedings{williams98_icslp,
author={Gethin Williams and Steve Renals},
title={{Confidence measures derived from an acceptor HMM}},
year=1998,
booktitle={Proc. 5th International Conference on Spoken Language Processing (ICSLP 1998)},
pages={paper 0644},
doi={10.21437/ICSLP.1998-434}
}