English syntactic reordering for English-Thai phrase-based statistical machine translation
Nawaphol Labutsri, Rapeeporn Chamchong, Richard Booth, Annupan Rodtook · 2009
In language pairs which have different wordorders, accuracy of translations in phrase-based statistical machine translation (SMT) systems will decrease. Syntactic reordering approaches can improve phrase-based SMT systems by reordering words in sentences to make word-orders of source language sentences similar to word-orders of target language sentences. This paper proposes reordering rules for an English-Thai phrase-based SMT system. Our reordering approach is the first that is tested in an English-Thai phrase-based SMT system. The reordering rules transform both training and test English sentences in a preprocessing step. After the preprocessing step, word-orders of English sentences are more similar to word-orders of Thai sentences. The reorder approach improves accuracy of English-Thai translation in the Moses phrase-based SMT system. In the system, the BLEU score increases clearly from 40.05 % to 57.45%. Key Words: reordering, English-Thai translation, phrase-based SMT