Discriminative language modeling for speech recognition with relevance information

Berlin Chen, Jia Wen Liu

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

5 Citations (Scopus)

Abstract

Discriminative language modeling (DLM) attempts to improve speech recognition performance by reranking the recognition hypotheses output from a baseline system. Most of the existing DLM methods assume that the reranking task can be treated as a linear discrimination problem and all testing utterances share the same parameter vector for reranking of hypotheses. However, the latter assumption sometimes results in a trained DLM model with weak generalizability and unsatisfactory performance. In view of this problem, we hence propose a relevance-based DLM (RDLM) framework that can efficiently infer the DLM model parameters of each testing utterance on-the-fly for better recognition performance. The structures and characteristics of the RDLM framework are extensively investigated, while the performance is thoroughly analyzed and verified by comparison with the existing DLM methods.

Original languageEnglish
Title of host publicationElectronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011
DOIs
Publication statusPublished - 2011 Nov 7
Event2011 12th IEEE International Conference on Multimedia and Expo, ICME 2011 - Barcelona, Spain
Duration: 2011 Jul 112011 Jul 15

Other

Other2011 12th IEEE International Conference on Multimedia and Expo, ICME 2011
CountrySpain
CityBarcelona
Period11/7/1111/7/15

Fingerprint

Speech recognition
Testing

Keywords

  • Discriminative Training
  • Language Modeling
  • Perceptron Method
  • Reranking
  • Speech Recognition

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications

Cite this

Chen, B., & Liu, J. W. (2011). Discriminative language modeling for speech recognition with relevance information. In Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011 [6012004] https://doi.org/10.1109/ICME.2011.6012004

Discriminative language modeling for speech recognition with relevance information. / Chen, Berlin; Liu, Jia Wen.

Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011. 2011. 6012004.

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

Chen, B & Liu, JW 2011, Discriminative language modeling for speech recognition with relevance information. in Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011., 6012004, 2011 12th IEEE International Conference on Multimedia and Expo, ICME 2011, Barcelona, Spain, 11/7/11. https://doi.org/10.1109/ICME.2011.6012004
Chen B, Liu JW. Discriminative language modeling for speech recognition with relevance information. In Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011. 2011. 6012004 https://doi.org/10.1109/ICME.2011.6012004
Chen, Berlin ; Liu, Jia Wen. / Discriminative language modeling for speech recognition with relevance information. Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011. 2011.
@inproceedings{ee72d6219b794558978d14ee5ae0ca35,
title = "Discriminative language modeling for speech recognition with relevance information",
abstract = "Discriminative language modeling (DLM) attempts to improve speech recognition performance by reranking the recognition hypotheses output from a baseline system. Most of the existing DLM methods assume that the reranking task can be treated as a linear discrimination problem and all testing utterances share the same parameter vector for reranking of hypotheses. However, the latter assumption sometimes results in a trained DLM model with weak generalizability and unsatisfactory performance. In view of this problem, we hence propose a relevance-based DLM (RDLM) framework that can efficiently infer the DLM model parameters of each testing utterance on-the-fly for better recognition performance. The structures and characteristics of the RDLM framework are extensively investigated, while the performance is thoroughly analyzed and verified by comparison with the existing DLM methods.",
keywords = "Discriminative Training, Language Modeling, Perceptron Method, Reranking, Speech Recognition",
author = "Berlin Chen and Liu, {Jia Wen}",
year = "2011",
month = "11",
day = "7",
doi = "10.1109/ICME.2011.6012004",
language = "English",
isbn = "9781612843490",
booktitle = "Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011",

}

TY - GEN

T1 - Discriminative language modeling for speech recognition with relevance information

AU - Chen, Berlin

AU - Liu, Jia Wen

PY - 2011/11/7

Y1 - 2011/11/7

N2 - Discriminative language modeling (DLM) attempts to improve speech recognition performance by reranking the recognition hypotheses output from a baseline system. Most of the existing DLM methods assume that the reranking task can be treated as a linear discrimination problem and all testing utterances share the same parameter vector for reranking of hypotheses. However, the latter assumption sometimes results in a trained DLM model with weak generalizability and unsatisfactory performance. In view of this problem, we hence propose a relevance-based DLM (RDLM) framework that can efficiently infer the DLM model parameters of each testing utterance on-the-fly for better recognition performance. The structures and characteristics of the RDLM framework are extensively investigated, while the performance is thoroughly analyzed and verified by comparison with the existing DLM methods.

AB - Discriminative language modeling (DLM) attempts to improve speech recognition performance by reranking the recognition hypotheses output from a baseline system. Most of the existing DLM methods assume that the reranking task can be treated as a linear discrimination problem and all testing utterances share the same parameter vector for reranking of hypotheses. However, the latter assumption sometimes results in a trained DLM model with weak generalizability and unsatisfactory performance. In view of this problem, we hence propose a relevance-based DLM (RDLM) framework that can efficiently infer the DLM model parameters of each testing utterance on-the-fly for better recognition performance. The structures and characteristics of the RDLM framework are extensively investigated, while the performance is thoroughly analyzed and verified by comparison with the existing DLM methods.

KW - Discriminative Training

KW - Language Modeling

KW - Perceptron Method

KW - Reranking

KW - Speech Recognition

UR - http://www.scopus.com/inward/record.url?scp=80155159173&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80155159173&partnerID=8YFLogxK

U2 - 10.1109/ICME.2011.6012004

DO - 10.1109/ICME.2011.6012004

M3 - Conference contribution

AN - SCOPUS:80155159173

SN - 9781612843490

BT - Electronic Proceedings of the 2011 IEEE International Conference on Multimedia and Expo, ICME 2011

ER -