Name
Title | ||
|---|---|---|
Structural Support Vector Machines for Log-Linear Approach in Statistical Machine Translation |
Abstract | ||
|---|---|---|
Minimum error rate training (MERT) is a widely used learn- ing method for statistical machine translation. In this pa- per, we present a SVM-based training method to enhance generalization ability. We extend MERT optimization by maximizing the margin between the reference and incor- rect translations under the L2-norm prior to avoid overfit- ting problem. Translation accuracy obtained by our proposed methods is more stable in various conditions than that ob- tained by MERT. Our experimental results on the French- English WMT08 shared task show that degrade of our pro- posed methods is smaller than that of MERT in case of small training data or out-of-domain test data. |
Year | Venue | Keywords |
|---|---|---|
2009 | IWSLT | support vector machine |
Field | DocType | Citations |
Training set,Pattern recognition,Computer science,Machine translation,Word error rate,Support vector machine,Artificial intelligence,Test data,Overfitting,Log-linear model,Machine learning | Conference | 2 |
PageRank | References | Authors |
0.36 | 18 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Katsuhiko Hayashi | 1 | 60 | 10.32 |
| Taro Watanabe | 2 | 572 | 36.86 |
| Hajime Tsukada | 3 | 449 | 29.46 |
| Hideki Isozaki | 4 | 934 | 64.50 |