Hope and Fear for Discriminative Training of Statistical Translation Models

David ChiangJMLR 2012 · OpenReview

Abstract

In machine translation, discriminative models have almost entirely supplanted the classical noisy-channel model, but are standardly trained using a method that is reliable only in low-dimensional spaces. Two strands of research have tried to adapt more scalable discriminative training methods to machine translation: the first uses log-linear probability models and either maximum likelihood or minimum risk, and the other uses linear models and large-margin methods. Here, we provide an overview of the latter. We compare several learning algorithms and describe in detail some novel extensions suited to properties of the translation task: no single correct output, a large space of structured outputs, and slow inference. We present experimental results on a large-scale Arabic-English translation task, demonstrating large gains in translation accuracy.
BibTeX
@article{chiang2012hope,
  author = {David Chiang},
  title = {{Hope and Fear for Discriminative Training of Statistical Translation Models}},
  journal = {J. Mach. Learn. Res.},
  year = {2012},
  url = {https://openreview.net/forum?id=sr8u87ip_iU},
}

Loading related content...

Figure