Minimum word error based discriminative training of language models

Jen Wei Kuo, Berlin Chen

Research output: Contribution to conferencePaper

12 Citations (Scopus)

Abstract

This paper considers discriminative training of language models for large vocabulary continuous speech recognition. The minimum word error (MWE) criterion was explored to make use of the word confusion information as well as the local lexical constraints inherent in the acoustic training corpus, in conjunction with those constraints obtained from the background text corpus, for properly guiding the speech recognizer to separate the correct hypothesis from the competing ones. The underlying characteristics of the MWE-based approach were extensively investigated, and its performance was verified by comparison with the conventional maximum likelihood (ML) approaches as well. The speech recognition experiments were performed on the broadcast news collected in Taiwan.

Original languageEnglish
Pages1277-1280
Number of pages4
Publication statusPublished - 2005 Dec 1
Event9th European Conference on Speech Communication and Technology - Lisbon, Portugal
Duration: 2005 Sep 42005 Sep 8

Other

Other9th European Conference on Speech Communication and Technology
CountryPortugal
CityLisbon
Period05/9/405/9/8

Fingerprint

Continuous speech recognition
Speech recognition
Maximum likelihood
Acoustics
Experiments

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Kuo, J. W., & Chen, B. (2005). Minimum word error based discriminative training of language models. 1277-1280. Paper presented at 9th European Conference on Speech Communication and Technology, Lisbon, Portugal.

Minimum word error based discriminative training of language models. / Kuo, Jen Wei; Chen, Berlin.

2005. 1277-1280 Paper presented at 9th European Conference on Speech Communication and Technology, Lisbon, Portugal.

Research output: Contribution to conferencePaper

Kuo, JW & Chen, B 2005, 'Minimum word error based discriminative training of language models', Paper presented at 9th European Conference on Speech Communication and Technology, Lisbon, Portugal, 05/9/4 - 05/9/8 pp. 1277-1280.
Kuo JW, Chen B. Minimum word error based discriminative training of language models. 2005. Paper presented at 9th European Conference on Speech Communication and Technology, Lisbon, Portugal.
Kuo, Jen Wei ; Chen, Berlin. / Minimum word error based discriminative training of language models. Paper presented at 9th European Conference on Speech Communication and Technology, Lisbon, Portugal.4 p.
@conference{c25ca8d905a0407182df70af6f6737e1,
title = "Minimum word error based discriminative training of language models",
abstract = "This paper considers discriminative training of language models for large vocabulary continuous speech recognition. The minimum word error (MWE) criterion was explored to make use of the word confusion information as well as the local lexical constraints inherent in the acoustic training corpus, in conjunction with those constraints obtained from the background text corpus, for properly guiding the speech recognizer to separate the correct hypothesis from the competing ones. The underlying characteristics of the MWE-based approach were extensively investigated, and its performance was verified by comparison with the conventional maximum likelihood (ML) approaches as well. The speech recognition experiments were performed on the broadcast news collected in Taiwan.",
author = "Kuo, {Jen Wei} and Berlin Chen",
year = "2005",
month = "12",
day = "1",
language = "English",
pages = "1277--1280",
note = "9th European Conference on Speech Communication and Technology ; Conference date: 04-09-2005 Through 08-09-2005",

}

TY - CONF

T1 - Minimum word error based discriminative training of language models

AU - Kuo, Jen Wei

AU - Chen, Berlin

PY - 2005/12/1

Y1 - 2005/12/1

N2 - This paper considers discriminative training of language models for large vocabulary continuous speech recognition. The minimum word error (MWE) criterion was explored to make use of the word confusion information as well as the local lexical constraints inherent in the acoustic training corpus, in conjunction with those constraints obtained from the background text corpus, for properly guiding the speech recognizer to separate the correct hypothesis from the competing ones. The underlying characteristics of the MWE-based approach were extensively investigated, and its performance was verified by comparison with the conventional maximum likelihood (ML) approaches as well. The speech recognition experiments were performed on the broadcast news collected in Taiwan.

AB - This paper considers discriminative training of language models for large vocabulary continuous speech recognition. The minimum word error (MWE) criterion was explored to make use of the word confusion information as well as the local lexical constraints inherent in the acoustic training corpus, in conjunction with those constraints obtained from the background text corpus, for properly guiding the speech recognizer to separate the correct hypothesis from the competing ones. The underlying characteristics of the MWE-based approach were extensively investigated, and its performance was verified by comparison with the conventional maximum likelihood (ML) approaches as well. The speech recognition experiments were performed on the broadcast news collected in Taiwan.

UR - http://www.scopus.com/inward/record.url?scp=33745224107&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33745224107&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:33745224107

SP - 1277

EP - 1280

ER -