A Direct Syntax-Driven Reordering Model for Phrase-Based Machine Translation

This paper presents a direct word reordering model with novel syntax-based features for statistical machine translation. Reordering models address the problem of reordering source language into the word order of the target language. IBM Models 3 through 5 have reordering components that use surface word information but very little context information to determine the traversal order of the source sentence. Since the late 1990s, phrase-based machine translation solves much of the local reorderings by using phrasal translations. The problem of longdistance reordering has become a central research topic in modeling distortions. We present a syntax driven maximum entropy reordering model that directly predicts the source traversal order and is able to model arbitrarily long distance word movement. We show that this model significantly improves machine translation quality.