ISCA Archive IWSLT 2009
ISCA Archive IWSLT 2009

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

Katsuhiko Hayashi, Taro Watanabe, Hajime Tsukada, Hideki Isozaki

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.


Cite as: Hayashi, K., Watanabe, T., Tsukada, H., Isozaki, H. (2009) Structural support vector machines for log-linear approach in statistical machine translation. Proc. International Workshop on Spoken Language Translation (IWSLT 2009), 144-151

@inproceedings{hayashi09_iwslt,
  author={Katsuhiko Hayashi and Taro Watanabe and Hajime Tsukada and Hideki Isozaki},
  title={{Structural support vector machines for log-linear approach in statistical machine translation}},
  year=2009,
  booktitle={Proc. International Workshop on Spoken Language Translation (IWSLT 2009)},
  pages={144--151}
}