In most phrase-based statistical machine translation (SMT) systems, the translation model relies on word alignment, which serves as a constraint for the subsequent building of a phrase table. Word alignment is usually inferred by GIZA++, which implements all the IBM models and HMM model in the framework of Expectation Maximum (EM). In this paper, we present a fully Bayesian inference for word alignment. Different from the EM approach, the Bayesian inference makes use of all possible parameter values rather than estimating a single parameter value, from which we expect a more robust inference. After inferring the word alignment, current SMT systems usually train the phrase table from Viterbi word alignment, which is prone to learn incorrect phrases due to the word alignment mistakes. To overcome this drawback, a new phrase extraction method is proposed based on multiple Gibbs samples from Bayesian inference for word alignment. Empirical results show promising improvements over baselines in alignment quality as well as the translation performance.
Zezhong LI
Ritsumeikan University
Hideto IKEDA
Ritsumeikan University
Junichi FUKUMOTO
Ritsumeikan University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Zezhong LI, Hideto IKEDA, Junichi FUKUMOTO, "Bayesian Word Alignment and Phrase Table Training for Statistical Machine Translation" in IEICE TRANSACTIONS on Information,
vol. E96-D, no. 7, pp. 1536-1543, July 2013, doi: 10.1587/transinf.E96.D.1536.
Abstract: In most phrase-based statistical machine translation (SMT) systems, the translation model relies on word alignment, which serves as a constraint for the subsequent building of a phrase table. Word alignment is usually inferred by GIZA++, which implements all the IBM models and HMM model in the framework of Expectation Maximum (EM). In this paper, we present a fully Bayesian inference for word alignment. Different from the EM approach, the Bayesian inference makes use of all possible parameter values rather than estimating a single parameter value, from which we expect a more robust inference. After inferring the word alignment, current SMT systems usually train the phrase table from Viterbi word alignment, which is prone to learn incorrect phrases due to the word alignment mistakes. To overcome this drawback, a new phrase extraction method is proposed based on multiple Gibbs samples from Bayesian inference for word alignment. Empirical results show promising improvements over baselines in alignment quality as well as the translation performance.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E96.D.1536/_p
Copy
@ARTICLE{e96-d_7_1536,
author={Zezhong LI, Hideto IKEDA, Junichi FUKUMOTO, },
journal={IEICE TRANSACTIONS on Information},
title={Bayesian Word Alignment and Phrase Table Training for Statistical Machine Translation},
year={2013},
volume={E96-D},
number={7},
pages={1536-1543},
abstract={In most phrase-based statistical machine translation (SMT) systems, the translation model relies on word alignment, which serves as a constraint for the subsequent building of a phrase table. Word alignment is usually inferred by GIZA++, which implements all the IBM models and HMM model in the framework of Expectation Maximum (EM). In this paper, we present a fully Bayesian inference for word alignment. Different from the EM approach, the Bayesian inference makes use of all possible parameter values rather than estimating a single parameter value, from which we expect a more robust inference. After inferring the word alignment, current SMT systems usually train the phrase table from Viterbi word alignment, which is prone to learn incorrect phrases due to the word alignment mistakes. To overcome this drawback, a new phrase extraction method is proposed based on multiple Gibbs samples from Bayesian inference for word alignment. Empirical results show promising improvements over baselines in alignment quality as well as the translation performance.},
keywords={},
doi={10.1587/transinf.E96.D.1536},
ISSN={1745-1361},
month={July},}
Copy
TY - JOUR
TI - Bayesian Word Alignment and Phrase Table Training for Statistical Machine Translation
T2 - IEICE TRANSACTIONS on Information
SP - 1536
EP - 1543
AU - Zezhong LI
AU - Hideto IKEDA
AU - Junichi FUKUMOTO
PY - 2013
DO - 10.1587/transinf.E96.D.1536
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E96-D
IS - 7
JA - IEICE TRANSACTIONS on Information
Y1 - July 2013
AB - In most phrase-based statistical machine translation (SMT) systems, the translation model relies on word alignment, which serves as a constraint for the subsequent building of a phrase table. Word alignment is usually inferred by GIZA++, which implements all the IBM models and HMM model in the framework of Expectation Maximum (EM). In this paper, we present a fully Bayesian inference for word alignment. Different from the EM approach, the Bayesian inference makes use of all possible parameter values rather than estimating a single parameter value, from which we expect a more robust inference. After inferring the word alignment, current SMT systems usually train the phrase table from Viterbi word alignment, which is prone to learn incorrect phrases due to the word alignment mistakes. To overcome this drawback, a new phrase extraction method is proposed based on multiple Gibbs samples from Bayesian inference for word alignment. Empirical results show promising improvements over baselines in alignment quality as well as the translation performance.
ER -