Word reordering approaches for Bangla-English statistical machine translation

Abstract

We apply several word reordering techniques to a Bangla-English Statistical Machine Translation (SMT) system. We evaluate the approaches through their impact on the BLEU score for the phrase-based Bangla-English SMT system. According to the experimental results, automatic reordering rules have the most significant impact on the BLEU score. We also provide a new test set with multiple references between Bangla and English for SMT evaluation purposes and evaluate the reordering approaches on the extended test set. © 2010 Springer-Verlag Berlin Heidelberg.