The USYD-JD Speech Translation System for IWSLT2021

Proceedings of the 18th International Conference on Spoken Language Translation, pages 182–191 Bangkok, Thailand (Online), August 5–6, 2021. ©2021 Association for Computational Linguistics 182 The USYD-JD Speech Translation System for IWSLT2021 Liang Ding The University of Sydney ldin3097@sydney.edu.au Di Wu∗ Peking University inbath@163.com Dacheng Tao JD Explore Academy, JD.com dacheng.tao@gmail.com Abstract This paper describes the University of Syd- ney & JD’s joint submission of the IWSLT 2021 low resource speech translation task. We participated in the Swahili→English direc- tion and got the best scareBLEU (25.3) score among all the participants. Our constrained system is based on a pipeline framework, i.e. ASR and NMT. We trained our models with the officially provided ASR and MT datasets. The ASR system is based on the open-sourced tool Kaldi and this work mainly explores how to make the most of the NMT models. To re- duce the punctuation errors generated by ASR model, we employ our previous work SlotRe- fine to train a punctuation correction model. To achieve better translation performance, we explored the most recent effective strategies, including back translation, knowledge distilla- tion, multi-feature reranking and transductive finetuning. For model structure, we tried auto- regressive and non-autoregressive models, re- spectively. In addition, we proposed two novel pre-train approaches, i.e. de-noising training and bidirectional training to fully exploit the data. Extensive experiments show that adding the above techniques consistently improves the BLEU scores, and the final submission sys- tem outperforms the baseline (Transformer en- semble model trained with the original paral- lel data) by approximately 10.8 BLEU score, achieving the SOTA performance. 1 Introduction Recent years have seen a surge of interest in speech translation (ST, Ney 1999) task, that translates the source-side speech to the target-side text directly. The ST task contains two major components, Au- tomatic Speech Recognition (ASR, Jelinek 1997) and Machine Translation (MT, Koehn 2009). In this year’s IWSLT low-resource speech translation ∗Work was done when Di Wu was visiting at JD. task, our USYD-JD translation team participated in the Swahili to English track. We break the speech translation task into “ASR→NMT” pipeline, and mainly focus on the NMT component. For model frameworks, we tried autoregressive neural machine translation, including Transformer- BASE and -BIG (Vaswani et al., 2017), and non- autoregressive translation models (Gu et al., 2018). Also, we employ our previous work SlotRefine (Wu et al., 2020a) to tackle the case and punctuation problems after ASR. To make the most of the par- allel and monolingual data, we proposed two pre- train strategies, i.e. BIDIRECTIONAL PRETRAIN- ING §2.2 and DENOISING PRETRAINING §2.3, and employed two data augmentation strategies, i.e. BIDIRECTIONAL SELF-TRAINING §2.5 and TAGGED BACK TRANSLATION §2.7. Where the data used for tagged back translation are care- fully selected with our proposed multi-feature in-domain selection approach in §2.6. For post finetune/ process, we employed TRANSDUCTIVE FINE-TUNE §2.8 and a simple postprocessing ap- proach §2.10. This paper is structured as follows: Section 2 de- scribes the major approaches we used. We present the data descriptions in Section 3. The experiments …

Content truncated to 500 words