Contextual maximum entropy model for edit disfluency detection of spontaneous speech

Jui Feng Yeh, Chung-Hsien Wu, Wei Yen Wu

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

Abstract

This study describes an approach to edit disfluency detection based on maximum entropy (ME) using contextual features for rich transcription of spontaneous speech. The contextual features contain word-level, chunk-level and sentence-level features for edit disfluency modeling. Due to the problem of data sparsity, word-level features are determined according to the taxonomy of the primary features of the words defined in Hownet. Chunk-level features are extracted based on mutual information of the words. Sentence-level feature are identified according to verbs and their corresponding features. The Improved Iterative Scaling (IIS) algorithm is employed to estimate the optimal weights in the maximum entropy models. Performance on edit disfluency detection and interruption point detection are conducted for evaluation. Experimental results show that the proposed method outperforms the DF-gram approach.

Original languageEnglish
Title of host publicationChinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings
Pages578-589
Number of pages12
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

Maximum Entropy
Entropy
Taxonomies
Transcription
Taxonomy
Mutual Information
Sparsity
Model
Scaling
Evaluation
Experimental Results
Modeling
Estimate
Speech

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Yeh, J. F., Wu, C-H., & Wu, W. Y. (2006). Contextual maximum entropy model for edit disfluency detection of spontaneous speech. In Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings (pp. 578-589). (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_60
Yeh, Jui Feng ; Wu, Chung-Hsien ; Wu, Wei Yen. / Contextual maximum entropy model for edit disfluency detection of spontaneous speech. Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. pp. 578-589 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{22a0a78abd214817921ad556a9ba02e5,
title = "Contextual maximum entropy model for edit disfluency detection of spontaneous speech",
abstract = "This study describes an approach to edit disfluency detection based on maximum entropy (ME) using contextual features for rich transcription of spontaneous speech. The contextual features contain word-level, chunk-level and sentence-level features for edit disfluency modeling. Due to the problem of data sparsity, word-level features are determined according to the taxonomy of the primary features of the words defined in Hownet. Chunk-level features are extracted based on mutual information of the words. Sentence-level feature are identified according to verbs and their corresponding features. The Improved Iterative Scaling (IIS) algorithm is employed to estimate the optimal weights in the maximum entropy models. Performance on edit disfluency detection and interruption point detection are conducted for evaluation. Experimental results show that the proposed method outperforms the DF-gram approach.",
author = "Yeh, {Jui Feng} and Chung-Hsien Wu and Wu, {Wei Yen}",
year = "2006",
month = "12",
day = "1",
doi = "10.1007/11939993_60",
language = "English",
isbn = "3540496653",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "578--589",
booktitle = "Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings",

}

Yeh, JF, Wu, C-H & Wu, WY 2006, Contextual maximum entropy model for edit disfluency detection of spontaneous speech. 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. 578-589, 5th International Symposium on Chinese Spoken Language Processing, ISCSLP 2006, Singapore, Singapore, 06-12-13. https://doi.org/10.1007/11939993_60

Contextual maximum entropy model for edit disfluency detection of spontaneous speech. / Yeh, Jui Feng; Wu, Chung-Hsien; Wu, Wei Yen.

Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. p. 578-589 (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

TY - GEN

T1 - Contextual maximum entropy model for edit disfluency detection of spontaneous speech

AU - Yeh, Jui Feng

AU - Wu, Chung-Hsien

AU - Wu, Wei Yen

PY - 2006/12/1

Y1 - 2006/12/1

N2 - This study describes an approach to edit disfluency detection based on maximum entropy (ME) using contextual features for rich transcription of spontaneous speech. The contextual features contain word-level, chunk-level and sentence-level features for edit disfluency modeling. Due to the problem of data sparsity, word-level features are determined according to the taxonomy of the primary features of the words defined in Hownet. Chunk-level features are extracted based on mutual information of the words. Sentence-level feature are identified according to verbs and their corresponding features. The Improved Iterative Scaling (IIS) algorithm is employed to estimate the optimal weights in the maximum entropy models. Performance on edit disfluency detection and interruption point detection are conducted for evaluation. Experimental results show that the proposed method outperforms the DF-gram approach.

AB - This study describes an approach to edit disfluency detection based on maximum entropy (ME) using contextual features for rich transcription of spontaneous speech. The contextual features contain word-level, chunk-level and sentence-level features for edit disfluency modeling. Due to the problem of data sparsity, word-level features are determined according to the taxonomy of the primary features of the words defined in Hownet. Chunk-level features are extracted based on mutual information of the words. Sentence-level feature are identified according to verbs and their corresponding features. The Improved Iterative Scaling (IIS) algorithm is employed to estimate the optimal weights in the maximum entropy models. Performance on edit disfluency detection and interruption point detection are conducted for evaluation. Experimental results show that the proposed method outperforms the DF-gram approach.

UR - http://www.scopus.com/inward/record.url?scp=77249125843&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77249125843&partnerID=8YFLogxK

U2 - 10.1007/11939993_60

DO - 10.1007/11939993_60

M3 - Conference contribution

SN - 3540496653

SN - 9783540496656

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

SP - 578

EP - 589

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

ER -

Yeh JF, Wu C-H, Wu WY. Contextual maximum entropy model for edit disfluency detection of spontaneous speech. In Chinese Spoken Language Processing - 5th International Symposium, ISCSLP 2006, Proceedings. 2006. p. 578-589. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/11939993_60