A Study of Reinforcement Learning for Neural Machine Translation
This work addresses the problem of improving NMT systems for translation tasks, but it is incremental as it builds on existing RL approaches with systematic comparisons and a new application to monolingual data.
The study tackled the challenge of unstable reinforcement learning (RL) training for neural machine translation (NMT) by systematically analyzing factors like baseline reward and reward shaping, and proposed a method to use RL with monolingual data, achieving competitive results including state-of-the-art performance on the WMT17 Chinese-English task.
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.