In the past researches, several kinds of information have been explored to assess the confidence measure or to select the confidence tag for a word/phrase. However, the contextual confidence information is little touched. In this paper, we propose a concept-based probabilistic verification model to integrate the contextual confidence information. In this model, a concept is verified not only according to its acoustic confidence measure but also according to neighboring concepts and their confidence levels. Experimental results show that the proposed model significantly outperforms the model using only confidence measures. The error rate of confidence tag is reduced from 17.7% to 15.12%, which corresponds to an error reduction rate of 14.5%.
Cite as: Lin, Y.-C., Wang, H.-M. (2001) Probabilistic concept verification for language understanding in spoken dialogue systems. Proc. 7th European Conference on Speech Communication and Technology (Eurospeech 2001), 1049-1052, doi: 10.21437/Eurospeech.2001-225
@inproceedings{lin01b_eurospeech,
author={Yi-Chung Lin and Huei-Ming Wang},
title={{Probabilistic concept verification for language understanding in spoken dialogue systems}},
year=2001,
booktitle={Proc. 7th European Conference on Speech Communication and Technology (Eurospeech 2001)},
pages={1049--1052},
doi={10.21437/Eurospeech.2001-225}
}