Computer Assisted Translation with Neural Quality Estimation and Automatic Post-Editing

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 2175–2186 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 2175 Computer Assisted Translation with Neural Quality Estimation and Automatic Post-Editing Ke Wang⇤, Jiayi Wang⇤, Niyu Ge, Yangbin Shi, Yu Zhao, Kai Fan† Alibaba Group Inc. {moyu.wk,joanne.wjy,niyu.ge,taiwu.syb}@alibaba-inc.com, kongyu@taobao.com, k.fan@alibaba-inc.com Abstract With the advent of neural machine transla- tion, there has been a marked shift towards leveraging and consuming the machine trans- lation results. However, the gap between ma- chine translation systems and human trans- lators needs to be manually closed by post- editing. In this paper, we propose an end-to- end deep learning framework of the quality es- timation and automatic post-editing of the ma- chine translation output. Our goal is to pro- vide error correction suggestions and to fur- ther relieve the burden of human translators through an interpretable model. To imitate the behavior of human translators, we design three efficient delegation modules – quality estima- tion, generative post-editing, and atomic oper- ation post-editing and construct a hierarchical model based on them. We examine this ap- proach with the English - German dataset from WMT 2017 APE shared task and our experi- mental results can achieve the state-of-the-art performance. We also verify that the certified translators can significantly expedite their post- editing processing with our model in human evaluation. 1 Introduction The explosive advances in the sequence to sequence model (Sutskever et al., 2014; Bahdanau et al., 2014; Vaswani et al., 2017) enable the deep learn- ing based neural machine translation (NMT) to approximate and even achieve the human parity in some specific language pairs and scenarios. Instead of translating from scratch by human translators, a new translation paradigm has emerged: computer assisted translation (CAT) system, which includes the machine translation and human post-editing. The post-editing is the process whereby humans amend machine-generated translations to achieve ⇤indicates equal contribution. †indicates corresponding author. an acceptable final product. Practically, the esti- mated average translation time can be reduced by 17.4% (from 1957.4 to 1617.7 seconds per text) (L¨aubli et al., 2013). However, utilizing NMT poses two key chal- lenges. First, the neural machine translation quality still continues to vary a great deal across different domains or genres, more or less in proportion to the availability of paralleled training corpora. Sec- ond, the zero tolerance policy is a common choice in the vast majority of important applications. For example, when business legal documents are trans- lated, even a single incorrect word could bring se- rious financial or property losses. Therefore, the subsequent human post-editing is indispensable in situations like this. Unfortunately, while NMT systems saves time by providing the preliminary translations, the time spent on error corrections by humans (L¨aubli et al., 2013) remains substantial to the extent that it offsets the efficiency gained by the NMT systems. In this paper, we explore automatic post-editing (APE) in the deep learning framework. Specifically, we adopt an imitation learning approach, where our model first screens the translation candidates by quality prediction and then …

Content truncated to 500 words