Combining Bilingual and Comparable Corpora for Low Resource Machine Translation

Statistical machine translation (SMT) performance suffers when models are trained on only small amounts of parallel data. The learned models typically have both low accuracy (incorrect translations and feature scores) and low coverage (high out-of-vocabulary rates). In this work, we use an additional data resource, comparable corpora, to improve both. Beginning with a small bitext and corresponding phrase-based SMT model, we improve coverage by using bilingual lexicon induction techniques to learn new translations from comparable corpora. Then, we supplement the model’s feature space with translation scores estimated over comparable corpora in order to improve accuracy. We observe improvements between 0.5 and 1.7 BLEU translating Tamil, Telugu, Bengali, Malayalam, Hindi, and Urdu into English.