On using entropy information to improve posterior probability-based confidence measures

Tzan Hwei Chen, Berlin Chen, Hsin Min Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

In this paper, we propose a novel approach that reduces the confidence error rate of traditional posterior probability-based confidence measures in large vocabulary continuous speech recognition systems. The method enhances the discriminability of confidence measures by applying entropy information to the posterior probability-based confidence measures of word hypotheses. The experiments conducted on the Chinese Mandarin broadcast news database MATBN show that entropy-based confidence measures outperform traditional posterior probability-based confidence measures. The relative reductions in the confidence error rate are 14.11% and 9.17% for experiments conducted on field reporter speech and interviewee speech, respectively.

Original languageEnglish
Title of host publicationChinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings
Pages454-463
Number of pages10
DOIs
Publication statusPublished - 2006 Dec 1
Event5th International Symposium on Chinese Spoken Language Processing, ISCSLP 2006 - Singapore, Singapore
Duration: 2006 Dec 132006 Dec 16

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4274 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other5th International Symposium on Chinese Spoken Language Processing, ISCSLP 2006
CountrySingapore
CitySingapore
Period06/12/1306/12/16

Fingerprint

Confidence Measure
Information Entropy
Posterior Probability
Entropy
Continuous speech recognition
Confidence
Error Rate
Experiments
Speech Recognition
Broadcast
Experiment

Keywords

  • Confidence measure
  • Continuous speech recognition
  • Entropy
  • Posterior probability

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Chen, T. H., Chen, B., & Wang, H. M. (2006). On using entropy information to improve posterior probability-based confidence measures. In Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings (pp. 454-463). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4274 LNAI). https://doi.org/10.1007/11939993_48

On using entropy information to improve posterior probability-based confidence measures. / Chen, Tzan Hwei; Chen, Berlin; Wang, Hsin Min.

Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. p. 454-463 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 4274 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Chen, TH, Chen, B & Wang, HM 2006, On using entropy information to improve posterior probability-based confidence measures. in Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 4274 LNAI, pp. 454-463, 5th International Symposium on Chinese Spoken Language Processing, ISCSLP 2006, Singapore, Singapore, 06/12/13. https://doi.org/10.1007/11939993_48
Chen TH, Chen B, Wang HM. On using entropy information to improve posterior probability-based confidence measures. In Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. p. 454-463. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/11939993_48
Chen, Tzan Hwei ; Chen, Berlin ; Wang, Hsin Min. / On using entropy information to improve posterior probability-based confidence measures. Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. pp. 454-463 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{2238ed3d98d44c2dadf4a28ce6152edd,
title = "On using entropy information to improve posterior probability-based confidence measures",
abstract = "In this paper, we propose a novel approach that reduces the confidence error rate of traditional posterior probability-based confidence measures in large vocabulary continuous speech recognition systems. The method enhances the discriminability of confidence measures by applying entropy information to the posterior probability-based confidence measures of word hypotheses. The experiments conducted on the Chinese Mandarin broadcast news database MATBN show that entropy-based confidence measures outperform traditional posterior probability-based confidence measures. The relative reductions in the confidence error rate are 14.11{\%} and 9.17{\%} for experiments conducted on field reporter speech and interviewee speech, respectively.",
keywords = "Confidence measure, Continuous speech recognition, Entropy, Posterior probability",
author = "Chen, {Tzan Hwei} and Berlin Chen and Wang, {Hsin Min}",
year = "2006",
month = "12",
day = "1",
doi = "10.1007/11939993_48",
language = "English",
isbn = "3540496653",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "454--463",
booktitle = "Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings",

}

TY - GEN

T1 - On using entropy information to improve posterior probability-based confidence measures

AU - Chen, Tzan Hwei

AU - Chen, Berlin

AU - Wang, Hsin Min

PY - 2006/12/1

Y1 - 2006/12/1

N2 - In this paper, we propose a novel approach that reduces the confidence error rate of traditional posterior probability-based confidence measures in large vocabulary continuous speech recognition systems. The method enhances the discriminability of confidence measures by applying entropy information to the posterior probability-based confidence measures of word hypotheses. The experiments conducted on the Chinese Mandarin broadcast news database MATBN show that entropy-based confidence measures outperform traditional posterior probability-based confidence measures. The relative reductions in the confidence error rate are 14.11% and 9.17% for experiments conducted on field reporter speech and interviewee speech, respectively.

AB - In this paper, we propose a novel approach that reduces the confidence error rate of traditional posterior probability-based confidence measures in large vocabulary continuous speech recognition systems. The method enhances the discriminability of confidence measures by applying entropy information to the posterior probability-based confidence measures of word hypotheses. The experiments conducted on the Chinese Mandarin broadcast news database MATBN show that entropy-based confidence measures outperform traditional posterior probability-based confidence measures. The relative reductions in the confidence error rate are 14.11% and 9.17% for experiments conducted on field reporter speech and interviewee speech, respectively.

KW - Confidence measure

KW - Continuous speech recognition

KW - Entropy

KW - Posterior probability

UR - http://www.scopus.com/inward/record.url?scp=77249152194&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77249152194&partnerID=8YFLogxK

U2 - 10.1007/11939993_48

DO - 10.1007/11939993_48

M3 - Conference contribution

AN - SCOPUS:77249152194

SN - 3540496653

SN - 9783540496656

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 454

EP - 463

BT - Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings

ER -