Structural support vector machines for log-linear approach in statistical machine translation

Katsuhiko Hayashi, Taro Watanabe, Hajime Tsukada, Hideki Isozaki


Abstract
Minimum error rate training (MERT) is a widely used learning method for statistical machine translation. In this paper, we present a SVM-based training method to enhance generalization ability. We extend MERT optimization by maximizing the margin between the reference and incorrect translations under the L2-norm prior to avoid overfitting problem. Translation accuracy obtained by our proposed methods is more stable in various conditions than that obtained by MERT. Our experimental results on the French-English WMT08 shared task show that degrade of our proposed methods is smaller than that of MERT in case of small training data or out-of-domain test data.
Anthology ID:
2009.iwslt-papers.3
Volume:
Proceedings of the 6th International Workshop on Spoken Language Translation: Papers
Month:
December 1-2
Year:
2009
Address:
Tokyo, Japan
Venue:
IWSLT
SIG:
SIGSLT
Publisher:
Note:
Pages:
144–151
Language:
URL:
https://aclanthology.org/2009.iwslt-papers.3
DOI:
Bibkey:
Cite (ACL):
Katsuhiko Hayashi, Taro Watanabe, Hajime Tsukada, and Hideki Isozaki. 2009. Structural support vector machines for log-linear approach in statistical machine translation. In Proceedings of the 6th International Workshop on Spoken Language Translation: Papers, pages 144–151, Tokyo, Japan.
Cite (Informal):
Structural support vector machines for log-linear approach in statistical machine translation (Hayashi et al., IWSLT 2009)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/2009.iwslt-papers.3.pdf
Presentation:
 2009.iwslt-papers.3.Presentation.pdf