In this paper, we present a recipe for building a good Arabic-English neural machine translation. We compare neural systems with traditional phrase-based systems using various parallel corpora including UN, ISI and Ummah. We also investigate the importance of special preprocessing of the Arabic script. The presented results are based on test sets from NIST MT 2005 and 2012. The best neural system produces a gain of +13 BLEU points compared to an equivalent simple phrase-based system in NIST MT12 test set. Unexpectedly, we find that tuning a model trained on the whole data using a small high quality corpus like Ummah gives a substantial improvement (+3 BLEU points). We also find that training a neural system with a small Arabic-English corpus is competitive to a traditional phrase-based system.
View on arXiv