Considerations in maximum mutual information and minimum classification error training for statistical machine translation

Ashish Venugopal, Stephan Vogel

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

7 Citations (Scopus)

Abstract

Discriminative training methods are used in statistical machine translation to effectively introduce and combine additional knowledge sources within the translation process. Although these methods are described in the accompanying literature and comparative studies are available for speech recognition, additional considerations are introduced when applying discriminative training to statistical machine translation. In this paper we pay special attention to the comparison and formalization of discriminative training criteria and their respective optimization methods with the goal of improving translation performance measured by the corpus level BLEU metric for a Viterbi beam based decoder. We frame this work within the current trends in discriminative training and present reproducible results that highlight the potential as well as shortcomings of N-Best list based discriminative training.

Original languageEnglish
Title of host publicationEuropean Association for Machine Translation, EAMT 2005 - 10th Annual Conference
Pages271-279
Number of pages9
Publication statusPublished - 1 Dec 2005
Externally publishedYes
Event10th Annual Conference on European Association for Machine Translation, EAMT 2005 - Budapest, Hungary
Duration: 30 May 200531 May 2005

Other

Other10th Annual Conference on European Association for Machine Translation, EAMT 2005
CountryHungary
CityBudapest
Period30/5/0531/5/05

Fingerprint

Speech recognition
Mutual Information
Statistical Machine Translation

ASJC Scopus subject areas

  • Language and Linguistics
  • Human-Computer Interaction
  • Software

Cite this

Venugopal, A., & Vogel, S. (2005). Considerations in maximum mutual information and minimum classification error training for statistical machine translation. In European Association for Machine Translation, EAMT 2005 - 10th Annual Conference (pp. 271-279)

Considerations in maximum mutual information and minimum classification error training for statistical machine translation. / Venugopal, Ashish; Vogel, Stephan.

European Association for Machine Translation, EAMT 2005 - 10th Annual Conference. 2005. p. 271-279.

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

Venugopal, A & Vogel, S 2005, Considerations in maximum mutual information and minimum classification error training for statistical machine translation. in European Association for Machine Translation, EAMT 2005 - 10th Annual Conference. pp. 271-279, 10th Annual Conference on European Association for Machine Translation, EAMT 2005, Budapest, Hungary, 30/5/05.
Venugopal A, Vogel S. Considerations in maximum mutual information and minimum classification error training for statistical machine translation. In European Association for Machine Translation, EAMT 2005 - 10th Annual Conference. 2005. p. 271-279
Venugopal, Ashish ; Vogel, Stephan. / Considerations in maximum mutual information and minimum classification error training for statistical machine translation. European Association for Machine Translation, EAMT 2005 - 10th Annual Conference. 2005. pp. 271-279
@inproceedings{1efaaf1455a540e395c990506db3d0d5,
title = "Considerations in maximum mutual information and minimum classification error training for statistical machine translation",
abstract = "Discriminative training methods are used in statistical machine translation to effectively introduce and combine additional knowledge sources within the translation process. Although these methods are described in the accompanying literature and comparative studies are available for speech recognition, additional considerations are introduced when applying discriminative training to statistical machine translation. In this paper we pay special attention to the comparison and formalization of discriminative training criteria and their respective optimization methods with the goal of improving translation performance measured by the corpus level BLEU metric for a Viterbi beam based decoder. We frame this work within the current trends in discriminative training and present reproducible results that highlight the potential as well as shortcomings of N-Best list based discriminative training.",
author = "Ashish Venugopal and Stephan Vogel",
year = "2005",
month = "12",
day = "1",
language = "English",
pages = "271--279",
booktitle = "European Association for Machine Translation, EAMT 2005 - 10th Annual Conference",

}

TY - GEN

T1 - Considerations in maximum mutual information and minimum classification error training for statistical machine translation

AU - Venugopal, Ashish

AU - Vogel, Stephan

PY - 2005/12/1

Y1 - 2005/12/1

N2 - Discriminative training methods are used in statistical machine translation to effectively introduce and combine additional knowledge sources within the translation process. Although these methods are described in the accompanying literature and comparative studies are available for speech recognition, additional considerations are introduced when applying discriminative training to statistical machine translation. In this paper we pay special attention to the comparison and formalization of discriminative training criteria and their respective optimization methods with the goal of improving translation performance measured by the corpus level BLEU metric for a Viterbi beam based decoder. We frame this work within the current trends in discriminative training and present reproducible results that highlight the potential as well as shortcomings of N-Best list based discriminative training.

AB - Discriminative training methods are used in statistical machine translation to effectively introduce and combine additional knowledge sources within the translation process. Although these methods are described in the accompanying literature and comparative studies are available for speech recognition, additional considerations are introduced when applying discriminative training to statistical machine translation. In this paper we pay special attention to the comparison and formalization of discriminative training criteria and their respective optimization methods with the goal of improving translation performance measured by the corpus level BLEU metric for a Viterbi beam based decoder. We frame this work within the current trends in discriminative training and present reproducible results that highlight the potential as well as shortcomings of N-Best list based discriminative training.

UR - http://www.scopus.com/inward/record.url?scp=84857582519&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84857582519&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84857582519

SP - 271

EP - 279

BT - European Association for Machine Translation, EAMT 2005 - 10th Annual Conference

ER -