Simpler and Faster Learning of Adaptive Policies for Simultaneous Translation

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1349–1354, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1349 Simpler and Faster Learning of Adaptive Policies for Simultaneous Translation Baigong Zheng 1,∗ Renjie Zheng 2,∗ Mingbo Ma 1 Liang Huang 1,2 1Baidu Research, Sunnyvale, CA, USA 2Oregon State University, Corvallis, OR, USA {baigongzheng, mingboma}@baidu.com zrenj11@gmail.com Abstract Simultaneous translation is widely useful but remains challenging. Previous work falls into two main categories: (a) fixed-latency poli- cies such as Ma et al. (2019) and (b) adap- tive policies such as Gu et al. (2017). The for- mer are simple and effective, but have to ag- gressively predict future content due to diverg- ing source-target word order; the latter do not anticipate, but suffer from unstable and ineffi- cient training. To combine the merits of both approaches, we propose a simple supervised- learning framework to learn an adaptive pol- icy from oracle READ/WRITE sequences gen- erated from parallel text. At each step, such an oracle sequence chooses to WRITE the next target word if the available source sen- tence context provides enough information to do so, otherwise READ the next source word. Experiments on German↔English show that our method, without retraining the underlying NMT model, can learn flexible policies with better BLEU scores and similar latencies com- pared to previous work. 1 Introduction Simultaneous translation outputs target words while the source sentence is being received, and is widely useful in international conferences, negoti- ations and press releases. However, although there is significant progress in machine translation (MT) recently, simultaneous machine translation is still one of the most challenging tasks. This is because it is hard to balance translation quality and latency, especially for the syntactically divergent language pairs, such as English and Japanese. Researchers previously study simultaneous translation as a part of real-time speech-to-speech translation system (Yarmohammadi et al., 2013; Bangalore et al., 2012; Fügen et al., 2007; Sridhar ∗ These authors contributed equally. et al., 2013; Jaitly et al., 2016; Graves et al., 2013). Recent simultaneous translation research focuses on obtaining a strategy, called a policy, to decide whether to wait for another source word (READ action) or emit a target word (WRITE action). The obtained policies fall into two main categories: (1) fixed-latency policies (Ma et al., 2019; Dalvi et al., 2018) and (2) context-dependent adaptive policies (Grissom II et al., 2014; Cho and Esipova, 2016; Gu et al., 2017; Alinejad et al., 2018; Arivazhagan et al., 2019; Zheng et al., 2019a). As an example of fixed-latency policies, wait-푘 (Ma et al., 2019) starts by waiting for the first few source words and then outputs one target word after receiving each new source word until the source sentence ends. It is easy to see that this kind of policy will inevitably need to guess the future content, which can often be incorrect. Thus, an adaptive policy (see Table 1 as an example), which can …

Content truncated to 500 words