Hidden) Conditional Random Fields Using Intermediate Classes for Statistical Machine Translation
Patrick Lehnen, Jan-Thorsten Peter, Joern Wuebker, Stephan Peitz, Hermann Ney · RWTH Publications (RWTH Aachen) · 2013
One of the major components of Statistical Machine Translation (SMT) are generative translation models.As in other fields, where the transition from generative to discriminative training resulted in higher performance, it seems likely that translation models should be trained in a discriminative way.But due to the nature of SMT with large vocabularies, hidden alignments, reordering, and large training corpora, the application of discriminative methods is only feasible when using effective speed up techniques.We will show that translation models trained with Conditional Random Fields (CRFs) using classes are useful in translation, even in addition to a strong baseline.Results with an independent CRF translation system and n-best list rescoring will be presented.To design the tandem of CRF translation model and a phrase based baseline we will evaluate two different ways of n-best list integrations.