A discriminative global training algorithm for statistical MT

Christoph Tillmann; Tong Zhang

doi:10.3115/1220175.1220266

A discriminative global training algorithm for statistical MT

Christoph Tillmann, Tong Zhang

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper presents a novel training algorithm for a linearly-scored block sequence translation model. The key component is a new procedure to directly optimize the global scoring function used by a SMT decoder. No translation, language, or distortion model probabilities are used as in earlier work on SMT. Therefore our method, which employs less domain specific knowledge, is both simpler and more extensible than previous approaches. Moreover, the training procedure treats the decoder as a black-box, and thus can be used to optimize any decoding scheme. The training algorithm is evaluated on a standard Arabic-English translation task.

Original language	English (US)
Title of host publication	COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference
Publisher	Association for Computational Linguistics (ACL)
Pages	721-728
Number of pages	8
ISBN (Print)	1932432655, 9781932432657
DOIs	https://doi.org/10.3115/1220175.1220266
State	Published - 2006
Externally published	Yes
Event	21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, COLING/ACL 2006 - Sydney, NSW, Australia Duration: Jul 17 2006 → Jul 21 2006

Publication series

Name	COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference
Volume	1

Other

Other	21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, COLING/ACL 2006
Country/Territory	Australia
City	Sydney, NSW
Period	7/17/06 → 7/21/06

ASJC Scopus subject areas

Language and Linguistics
Linguistics and Language

Online availability

10.3115/1220175.1220266

Library availability

Discover UIUC Full Text

Cite this

Tillmann, C., & Zhang, T. (2006). A discriminative global training algorithm for statistical MT. In COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (pp. 721-728). (COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference; Vol. 1). Association for Computational Linguistics (ACL). https://doi.org/10.3115/1220175.1220266

A discriminative global training algorithm for statistical MT. / Tillmann, Christoph; Zhang, Tong.
COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference. Association for Computational Linguistics (ACL), 2006. p. 721-728 (COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference; Vol. 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Tillmann, C & Zhang, T 2006, A discriminative global training algorithm for statistical MT. in COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference. COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference, vol. 1, Association for Computational Linguistics (ACL), pp. 721-728, 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, COLING/ACL 2006, Sydney, NSW, Australia, 7/17/06. https://doi.org/10.3115/1220175.1220266

Tillmann C, Zhang T. A discriminative global training algorithm for statistical MT. In COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference. Association for Computational Linguistics (ACL). 2006. p. 721-728. (COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference). doi: 10.3115/1220175.1220266

Tillmann, Christoph ; Zhang, Tong. / A discriminative global training algorithm for statistical MT. COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference. Association for Computational Linguistics (ACL), 2006. pp. 721-728 (COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference).

@inproceedings{e824ff04ac4f4a0a8349b4e110a09756,

title = "A discriminative global training algorithm for statistical MT",

abstract = "This paper presents a novel training algorithm for a linearly-scored block sequence translation model. The key component is a new procedure to directly optimize the global scoring function used by a SMT decoder. No translation, language, or distortion model probabilities are used as in earlier work on SMT. Therefore our method, which employs less domain specific knowledge, is both simpler and more extensible than previous approaches. Moreover, the training procedure treats the decoder as a black-box, and thus can be used to optimize any decoding scheme. The training algorithm is evaluated on a standard Arabic-English translation task.",

author = "Christoph Tillmann and Tong Zhang",

year = "2006",

doi = "10.3115/1220175.1220266",

language = "English (US)",

isbn = "1932432655",

series = "COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference",

publisher = "Association for Computational Linguistics (ACL)",

pages = "721--728",

booktitle = "COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference",

address = "United States",

note = "21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, COLING/ACL 2006 ; Conference date: 17-07-2006 Through 21-07-2006",

}

TY - GEN

T1 - A discriminative global training algorithm for statistical MT

AU - Tillmann, Christoph

AU - Zhang, Tong

PY - 2006

Y1 - 2006

N2 - This paper presents a novel training algorithm for a linearly-scored block sequence translation model. The key component is a new procedure to directly optimize the global scoring function used by a SMT decoder. No translation, language, or distortion model probabilities are used as in earlier work on SMT. Therefore our method, which employs less domain specific knowledge, is both simpler and more extensible than previous approaches. Moreover, the training procedure treats the decoder as a black-box, and thus can be used to optimize any decoding scheme. The training algorithm is evaluated on a standard Arabic-English translation task.

AB - This paper presents a novel training algorithm for a linearly-scored block sequence translation model. The key component is a new procedure to directly optimize the global scoring function used by a SMT decoder. No translation, language, or distortion model probabilities are used as in earlier work on SMT. Therefore our method, which employs less domain specific knowledge, is both simpler and more extensible than previous approaches. Moreover, the training procedure treats the decoder as a black-box, and thus can be used to optimize any decoding scheme. The training algorithm is evaluated on a standard Arabic-English translation task.

UR - http://www.scopus.com/inward/record.url?scp=84860527068&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84860527068&partnerID=8YFLogxK

U2 - 10.3115/1220175.1220266

DO - 10.3115/1220175.1220266

M3 - Conference contribution

AN - SCOPUS:84860527068

SN - 1932432655

SN - 9781932432657

T3 - COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference

SP - 721

EP - 728

BT - COLING/ACL 2006 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference

PB - Association for Computational Linguistics (ACL)

T2 - 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, COLING/ACL 2006

Y2 - 17 July 2006 through 21 July 2006

ER -

A discriminative global training algorithm for statistical MT

Abstract

Publication series

Other

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this