Source reordering using MaxEnt classifiers and supertags

Source language reordering can be seen as the preprocessing task of permuting the order of the source words in such a way that the resulting permutation allows as monotone a translation process as possible. We explore a simple but effective source reordering algorithm that works as a cascade of source string transforms, each consisting of swapping the positions of a single pair of adjacent words in order to unfold a candidate pair of crossing alignments. The decision to swap a pair of words is modelled as a binary classification task formulated as a log-linear model and trained under maximum entropy (MaxEnt). We experiment with features that consist of the local neighborhood of both words as well as lexico-syntactic representations known as supertags. Our experiments on the English-to-Dutch EuroParl translation task show that the cascaded alignment unfolding slightly improves the performance of a state-of-the-art phrase translation system that uses distance-based and lexicalized block-oriented reordering.