QCRI Machine Translation Systems for IWSLT 16

TL;DR

This paper compares phrase-based and neural machine translation systems for Arabic-English, demonstrating that NMT with ensemble and fine-tuning significantly outperforms traditional phrase-based models in the IWSLT 2016 evaluation.

Contribution

It presents a comprehensive comparison showing NMT's superiority over phrase-based systems for Arabic-English translation in a competitive setting.

Findings

NMT outperforms phrase-based models by 2 BLEU points in Arabic->English.

Ensemble and fine-tuning improve NMT performance.

System combination enhances translation quality.

Abstract

This paper describes QCRI's machine translation systems for the IWSLT 2016 evaluation campaign. We participated in the Arabic->English and English->Arabic tracks. We built both Phrase-based and Neural machine translation models, in an effort to probe whether the newly emerged NMT framework surpasses the traditional phrase-based systems in Arabic-English language pairs. We trained a very strong phrase-based system including, a big language model, the Operation Sequence Model, Neural Network Joint Model and Class-based models along with different domain adaptation techniques such as MML filtering, mixture modeling and using fine tuning over NNJM model. However, a Neural MT system, trained by stacking data from different genres through fine-tuning, and applying ensemble over 8 models, beat our very strong phrase-based system by a significant 2 BLEU points margin in Arabic->English direction. We did not obtain similar gains in the other direction but were still able to outperform the phrase-based system. We also applied system combination on phrase-based and NMT outputs.