IEICE TRANSACTIONS on Information
Word Error Rate Minimization Using an Integrated Confidence Measure
Summary :
This paper describes a new criterion for speech recognition using an integrated confidence measure to minimize the word error rate (WER). The conventional criteria for WER minimization obtain the expected WER of a sentence hypothesis merely by comparing it with other hypotheses in an n-best list. The proposed criterion estimates the expected WER by using an integrated confidence measure with word posterior probabilities for a given acoustic input. The integrated confidence measure, which is implemented as a classifier based on maximum entropy (ME) modeling or support vector machines (SVMs), is used to acquire probabilities reflecting whether the word hypotheses are correct. The classifier is comprised of a variety of confidence measures and can deal with a temporal sequence of them to attain a more reliable confidence. Our proposed criterion for minimizing WER achieved a WER of 9.8% and a 3.9% reduction, relative to conventional n-best rescoring methods in transcribing Japanese broadcast news in various environments such as under noisy field and spontaneous speech conditions.
- Publication
- IEICE TRANSACTIONS on Information Vol.E90-D No.5 pp.835-843
- Publication Date
- 2007/05/01
- Publicized
- Online ISSN
- 1745-1361
- DOI
- 10.1093/ietisy/e90-d.5.835
- Type of Manuscript
- PAPER
- Category
- Speech and Hearing
Authors
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Akio KOBAYASHI, Kazuo ONOE, Shinichi HOMMA, Shoei SATO, Toru IMAI, "Word Error Rate Minimization Using an Integrated Confidence Measure" in IEICE TRANSACTIONS on Information,
vol. E90-D, no. 5, pp. 835-843, May 2007, doi: 10.1093/ietisy/e90-d.5.835.
Abstract: This paper describes a new criterion for speech recognition using an integrated confidence measure to minimize the word error rate (WER). The conventional criteria for WER minimization obtain the expected WER of a sentence hypothesis merely by comparing it with other hypotheses in an n-best list. The proposed criterion estimates the expected WER by using an integrated confidence measure with word posterior probabilities for a given acoustic input. The integrated confidence measure, which is implemented as a classifier based on maximum entropy (ME) modeling or support vector machines (SVMs), is used to acquire probabilities reflecting whether the word hypotheses are correct. The classifier is comprised of a variety of confidence measures and can deal with a temporal sequence of them to attain a more reliable confidence. Our proposed criterion for minimizing WER achieved a WER of 9.8% and a 3.9% reduction, relative to conventional n-best rescoring methods in transcribing Japanese broadcast news in various environments such as under noisy field and spontaneous speech conditions.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e90-d.5.835/_p
Copy
@ARTICLE{e90-d_5_835,
author={Akio KOBAYASHI, Kazuo ONOE, Shinichi HOMMA, Shoei SATO, Toru IMAI, },
journal={IEICE TRANSACTIONS on Information},
title={Word Error Rate Minimization Using an Integrated Confidence Measure},
year={2007},
volume={E90-D},
number={5},
pages={835-843},
abstract={This paper describes a new criterion for speech recognition using an integrated confidence measure to minimize the word error rate (WER). The conventional criteria for WER minimization obtain the expected WER of a sentence hypothesis merely by comparing it with other hypotheses in an n-best list. The proposed criterion estimates the expected WER by using an integrated confidence measure with word posterior probabilities for a given acoustic input. The integrated confidence measure, which is implemented as a classifier based on maximum entropy (ME) modeling or support vector machines (SVMs), is used to acquire probabilities reflecting whether the word hypotheses are correct. The classifier is comprised of a variety of confidence measures and can deal with a temporal sequence of them to attain a more reliable confidence. Our proposed criterion for minimizing WER achieved a WER of 9.8% and a 3.9% reduction, relative to conventional n-best rescoring methods in transcribing Japanese broadcast news in various environments such as under noisy field and spontaneous speech conditions.},
keywords={},
doi={10.1093/ietisy/e90-d.5.835},
ISSN={1745-1361},
month={May},}
Copy
TY - JOUR
TI - Word Error Rate Minimization Using an Integrated Confidence Measure
T2 - IEICE TRANSACTIONS on Information
SP - 835
EP - 843
AU - Akio KOBAYASHI
AU - Kazuo ONOE
AU - Shinichi HOMMA
AU - Shoei SATO
AU - Toru IMAI
PY - 2007
DO - 10.1093/ietisy/e90-d.5.835
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E90-D
IS - 5
JA - IEICE TRANSACTIONS on Information
Y1 - May 2007
AB - This paper describes a new criterion for speech recognition using an integrated confidence measure to minimize the word error rate (WER). The conventional criteria for WER minimization obtain the expected WER of a sentence hypothesis merely by comparing it with other hypotheses in an n-best list. The proposed criterion estimates the expected WER by using an integrated confidence measure with word posterior probabilities for a given acoustic input. The integrated confidence measure, which is implemented as a classifier based on maximum entropy (ME) modeling or support vector machines (SVMs), is used to acquire probabilities reflecting whether the word hypotheses are correct. The classifier is comprised of a variety of confidence measures and can deal with a temporal sequence of them to attain a more reliable confidence. Our proposed criterion for minimizing WER achieved a WER of 9.8% and a 3.9% reduction, relative to conventional n-best rescoring methods in transcribing Japanese broadcast news in various environments such as under noisy field and spontaneous speech conditions.
ER -
English
