A Study of Reinforcement Learning for Neural Machine Translation

TL;DR

This paper systematically studies reinforcement learning for neural machine translation, addressing training challenges, comparing key factors, and proposing a new method to enhance performance using monolingual data, achieving state-of-the-art results.

Contribution

It provides a comprehensive analysis of RL training for NMT, introduces a novel approach leveraging monolingual data, and demonstrates improved translation performance on multiple benchmarks.

Findings

RL training stability varies with baseline reward and reward shaping.

The proposed method effectively utilizes monolingual data to boost NMT performance.

Achieved state-of-the-art results on WMT17 Chinese-English translation.

Abstract

Recent studies have shown that reinforcement learning (RL) is an effective approach for improving the performance of neural machine translation (NMT) system. However, due to its instability, successfully RL training is challenging, especially in real-world systems where deep models and large datasets are leveraged. In this paper, taking several large-scale translation tasks as testbeds, we conduct a systematic study on how to train better NMT models using reinforcement learning. We provide a comprehensive comparison of several important factors (e.g., baseline reward, reward shaping) in RL training. Furthermore, to fill in the gap that it remains unclear whether RL is still beneficial when monolingual data is used, we propose a new method to leverage RL to further boost the performance of NMT systems trained with source/target monolingual data. By integrating all our findings, we obtain competitive results on WMT14 English- German, WMT17 English-Chinese, and WMT17 Chinese-English translation tasks, especially setting a state-of-the-art performance on WMT17 Chinese-English translation task.