Dual-Path phrase-based statistical machine translation

Preceding a phrase-based statistical machine translation (PSMT) system by a syntactically-informed reordering preprocessing step has been shown to improve overall translation performance compared to a baseline PSMT system. However, the improvement is not seen for every sentence. We use a lattice input to a PSMT system in order to translate simultaneously across both original and reordered versions of a sentence, and include a number of confidence features to support the system in choosing on a sentence-by-sentence basis whether to use the reordering process. In German-to-English translation, our best system achieves a BLEU score of 21.39, an improvement of 0.62.