21 Mar 2011

Time: Friday, March 25th, 2011 11:00 a.m.

Place: TASC1 9204 West

Title: Application of Tree Transducers in Statistical Machine Translation

Abstract:

The need to incorporate the syntactic structure of sentences in machine translation has led to introducing syntax-based systems. As a result, tree transducers have gained significant interest from the machine translation community in recent years. Tree transducers can neatly model most of the current syntax-based machine translation models. They provide a framework within which generic algorithms can be designed and implemented. We survey the application of tree transducers in machine translation. More specifically, we explore different approaches to constructing tree transducers from a tree-bank as well as methods for training and decoding tree transducers.