Regularization and Search for Minimum Error Rate Training
Cached
Download Links
by
Daniel Cer
,
Daniel Jurafsky
,
Christopher D. Manning
| Citations: | 7 - 0 self |
BibTeX
@MISC{Cer_regularizationand,
author = {Daniel Cer and Daniel Jurafsky and Christopher D. Manning},
title = {Regularization and Search for Minimum Error Rate Training},
year = {}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
Minimum error rate training (MERT) is a widely used learning procedure for statistical machine translation models. We contrast three search strategies for MERT: Powell’s method, the variant of coordinate descent found in the Moses MERT utility, and a novel stochastic method. It is shown that the stochastic method obtains test set gains of +0.98 BLEU on MT03 and +0.61 BLEU on MT05. We also present a method for regularizing the MERT objective that achieves statistically significant gains when combined with both Powell’s method and coordinate descent. 1
