Word topical mixture models for dynamic language model adaptation

Hsuan Sheng Chiu, Berlin Chen

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

19 Citations (Scopus)

Abstract

This paper considers dynamic language model adaptation for Mandarin broadcast news recognition. A word topical mixture model (TMM) is proposed to explore the co-occurrence relationship between words, as well as the long-span latent topical information, for language model adaptation. The search history is modeled as a composite word TMM model for predicting the decoded word. The underlying characteristics and different kinds of model structures were extensively investigated, while the performance of word TMM was analyzed and verified by comparison with the conventional probabilistic latent semantic analysis-based language model (PLSALM) and trigger-based language model (TBLM) adaptation approaches. The large vocabulary continuous speech recognition (LVCSR) experiments were conducted on the Mandarin broadcast news collected in Taiwan. Very promising results in perplexity as well as character error rate reductions were initially obtained.

Original languageEnglish
Title of host publication2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07
PagesIV169-IV172
DOIs
Publication statusPublished - 2007 Aug 6
Event2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 - Honolulu, HI, United States
Duration: 2007 Apr 152007 Apr 20

Publication series

NameICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume4
ISSN (Print)1520-6149

Other

Other2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07
CountryUnited States
CityHonolulu, HI
Period07/4/1507/4/20

Fingerprint

Continuous speech recognition
Model structures
Semantics
Composite materials
Experiments

Keywords

  • Language model adaptation
  • Probabilistic latent semantic analysis
  • Speech recognition
  • Trigger-based language model
  • Word topical mixture model

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Cite this

Chiu, H. S., & Chen, B. (2007). Word topical mixture models for dynamic language model adaptation. In 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07 (pp. IV169-IV172). [4218064] (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 4). https://doi.org/10.1109/ICASSP.2007.367190

Word topical mixture models for dynamic language model adaptation. / Chiu, Hsuan Sheng; Chen, Berlin.

2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07. 2007. p. IV169-IV172 4218064 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 4).

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

Chiu, HS & Chen, B 2007, Word topical mixture models for dynamic language model adaptation. in 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07., 4218064, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 4, pp. IV169-IV172, 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07, Honolulu, HI, United States, 07/4/15. https://doi.org/10.1109/ICASSP.2007.367190
Chiu HS, Chen B. Word topical mixture models for dynamic language model adaptation. In 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07. 2007. p. IV169-IV172. 4218064. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). https://doi.org/10.1109/ICASSP.2007.367190
Chiu, Hsuan Sheng ; Chen, Berlin. / Word topical mixture models for dynamic language model adaptation. 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07. 2007. pp. IV169-IV172 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).
@inproceedings{915a2a5b748d4f29b52fe3cc694419d0,
title = "Word topical mixture models for dynamic language model adaptation",
abstract = "This paper considers dynamic language model adaptation for Mandarin broadcast news recognition. A word topical mixture model (TMM) is proposed to explore the co-occurrence relationship between words, as well as the long-span latent topical information, for language model adaptation. The search history is modeled as a composite word TMM model for predicting the decoded word. The underlying characteristics and different kinds of model structures were extensively investigated, while the performance of word TMM was analyzed and verified by comparison with the conventional probabilistic latent semantic analysis-based language model (PLSALM) and trigger-based language model (TBLM) adaptation approaches. The large vocabulary continuous speech recognition (LVCSR) experiments were conducted on the Mandarin broadcast news collected in Taiwan. Very promising results in perplexity as well as character error rate reductions were initially obtained.",
keywords = "Language model adaptation, Probabilistic latent semantic analysis, Speech recognition, Trigger-based language model, Word topical mixture model",
author = "Chiu, {Hsuan Sheng} and Berlin Chen",
year = "2007",
month = "8",
day = "6",
doi = "10.1109/ICASSP.2007.367190",
language = "English",
isbn = "1424407281",
series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
pages = "IV169--IV172",
booktitle = "2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07",

}

TY - GEN

T1 - Word topical mixture models for dynamic language model adaptation

AU - Chiu, Hsuan Sheng

AU - Chen, Berlin

PY - 2007/8/6

Y1 - 2007/8/6

N2 - This paper considers dynamic language model adaptation for Mandarin broadcast news recognition. A word topical mixture model (TMM) is proposed to explore the co-occurrence relationship between words, as well as the long-span latent topical information, for language model adaptation. The search history is modeled as a composite word TMM model for predicting the decoded word. The underlying characteristics and different kinds of model structures were extensively investigated, while the performance of word TMM was analyzed and verified by comparison with the conventional probabilistic latent semantic analysis-based language model (PLSALM) and trigger-based language model (TBLM) adaptation approaches. The large vocabulary continuous speech recognition (LVCSR) experiments were conducted on the Mandarin broadcast news collected in Taiwan. Very promising results in perplexity as well as character error rate reductions were initially obtained.

AB - This paper considers dynamic language model adaptation for Mandarin broadcast news recognition. A word topical mixture model (TMM) is proposed to explore the co-occurrence relationship between words, as well as the long-span latent topical information, for language model adaptation. The search history is modeled as a composite word TMM model for predicting the decoded word. The underlying characteristics and different kinds of model structures were extensively investigated, while the performance of word TMM was analyzed and verified by comparison with the conventional probabilistic latent semantic analysis-based language model (PLSALM) and trigger-based language model (TBLM) adaptation approaches. The large vocabulary continuous speech recognition (LVCSR) experiments were conducted on the Mandarin broadcast news collected in Taiwan. Very promising results in perplexity as well as character error rate reductions were initially obtained.

KW - Language model adaptation

KW - Probabilistic latent semantic analysis

KW - Speech recognition

KW - Trigger-based language model

KW - Word topical mixture model

UR - http://www.scopus.com/inward/record.url?scp=34547505668&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34547505668&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2007.367190

DO - 10.1109/ICASSP.2007.367190

M3 - Conference contribution

AN - SCOPUS:34547505668

SN - 1424407281

SN - 9781424407286

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - IV169-IV172

BT - 2007 IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP '07

ER -