Beyond log-linear models: Boosted minimum error rate training for nbest re-ranking (2008)
| Venue: | In Proc. ACL-08: HLT, Short Papers |
| Citations: | 4 - 0 self |
BibTeX
@INPROCEEDINGS{Duh08beyondlog-linear,
author = {Kevin Duh and Katrin Kirchhoff},
title = {Beyond log-linear models: Boosted minimum error rate training for nbest re-ranking},
booktitle = {In Proc. ACL-08: HLT, Short Papers},
year = {2008}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Current re-ranking algorithms for machine translation rely on log-linear models, which have the potential problem of underfitting the training data. We present BoostedMERT, a novel boosting algorithm that uses Minimum Error Rate Training (MERT) as a weak learner and builds a re-ranker far more expressive than log-linear models. BoostedMERT is easy to implement, inherits the efficient optimization properties of MERT, and can quickly boost the BLEU score on N-best re-ranking tasks. In this paper, we describe the general algorithm and present preliminary results on the IWSLT 2007 Arabic-English task. 1
Citations
| 561 | Y: Improved boosting algorithms using confidence-rated predictions - RE, Singer - 1999 |
| 383 | An efficient boosting algorithm for combining preferences - Freund, Iyer, et al. - 1998 |
| 93 | A smorgasbord of features for statistical machine translation - Och, Gildea, et al. - 2004 |
| 43 | Discriminative reranking for machine translation - Shen, Sarkar, et al. - 2004 |
| 38 | 2006. Minimum risk annealing for training log-linear models - Smith, Eisner |
| 16 | Boosting-based parse reranking with subtree features - Kudo, Suzuki, et al. - 2005 |
| 4 | Boosting as gradient descent - Mason, Baxter, et al. - 2000 |
| 1 | Discriminative reranking for natural langauge parsing - Collins, Koo - 2005 |
| 1 | The UW machine translation system for IWSLT 2007 - Kirchhoff, Yang - 2007 |
| 1 | Training and evaluating error minimization rules for SMT - Venugopal, Zollmann, et al. - 2005 |
