The Volctrans Neural Speech Translation System for IWSLT 2021

Proceedings of the 18th International Conference on Spoken Language Translation, pages 64–74 Bangkok, Thailand (Online), August 5–6, 2021. ©2021 Association for Computational Linguistics 64 The Volctrans Neural Speech Translation System for IWSLT 2021 Chengqi Zhao Zhicheng Liu Jian Tong Tao Wang Mingxuan Wang Rong Ye Qianqian Dong Jun Cao Lei Li ByteDance AI Lab {zhaochengqi.d,liuzhicheng.lzc,tongjian,wangtao.960826 wangmingxuan.89,yerong,dongqianqian caojun.sh,lileilab}@bytedance.com Abstract This paper describes the systems submitted to IWSLT 2021 by the Volctrans team. We partic- ipate in the offline speech translation and text- to-text simultaneous translation tracks. For of- fline speech translation, our best end-to-end model achieves 7.9 BLEU improvements over the benchmark on the MuST-C test set and is even approaching the results of a strong cas- cade solution. For text-to-text simultaneous translation, we explore the best practice to op- timize the wait-k model. As a result, our final submitted systems exceed the benchmark at around 7 BLEU on the same latency regime. We release our code and model to facilitate both future research works and industrial ap- plications1. 1 Introduction This paper describes the neural speech translation systems submitted to IWSLT 2021 by the Volctrans team (also known as the team from ByteDance AI Lab), including cascade and end-to-end speech translation (ST) systems for the offline ST track and a simultaneous neural machine translation (NMT) system. We aim at finding the best practice for these two tracks. For offline ST, the cascaded system often out- performs the fully end-to-end approach. Recent studies on the fully end-to-end approaches obtain promising results and attract a lot of interest. Last year’s results have shown that an end-to-end model achieves an even better performance (Ansari et al., 2020) compared with the cascaded competitors. However, they introduce pre-training (Bansal et al., 2019; Stoian et al., 2020; Wang et al., 2020; Aline- jad and Sarkar, 2020) and data augmentation tech- niques (Jia et al., 2019; Pino et al., 2020) to end-to- end models, while the cascaded is not that strong 1Code and models are available at //github.com/bytedance/neurst/tree/https: master/examples/iwslt21 enough. Hence, in this paper, we would like to optimize the speech translation model in two as- pects. First, we are devoted to building a strong cascade competitor and learns the best practice from WMT evaluation campaigns (Li et al., 2019; Wu et al., 2020), such as back translation (Sennrich et al., 2016a) and ensemble. Second, we explore various self-supervised learning methods and in- troduce as much semi-supervised data as possible towards finding the best practice of training end- to-end ST models. In our settings, ASR data, MT data, and monolingual text data are all considered in a progressively training framework. The results are very promising, and the final performance on the MuST-C test set surpasses the end-to-end base- line by 7.9 BLUE scores, while it is still lagging behind our cascade model by 1.5 BLUE scores. It is not surprising since some well-optimized meth- ods for MT can not be easily used on ST, such as back translation. However, our experience shows that the external data can …

Content truncated to 500 words