Towards User-Driven Neural Machine Translation

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 4008–4018 August 1–6, 2021. ©2021 Association for Computational Linguistics 4008 Towards User-Driven Neural Machine Translation Huan Lin1,2 Liang Yao3 Baosong Yang3 Dayiheng Liu3 Haibo Zhang3 Weihua Luo3 Degen Huang4 Jinsong Su1,2,5∗ 1School of Informatics, Xiamen University 2Institute of Artificial Intelligence, Xiamen University 3Alibaba Group 4Dalian University of Technology 5Pengcheng Lab, Shenzhen huanlin@stu.xmu.edu.cn {yaoliang.yl,yangbaosong.ybs,liudayiheng.ldyh,zhanhui.zhb}@alibaba-inc.com weihua.luowh@alibaba-inc.com huangdg@dlut.edu.cn jssu@xmu.edu.cn Abstract A good translation should not only translate the original content semantically, but also in- carnate personal traits of the original text. For a real-world neural machine translation (NMT) system, these user traits (e.g., topic preference, stylistic characteristics and expression habits) can be preserved in user behavior (e.g., histor- ical inputs). However, current NMT systems marginally consider the user behavior due to: 1) the difficulty of modeling user portraits in zero-shot scenarios, and 2) the lack of user- behavior annotated parallel dataset. To fill this gap, we introduce a novel framework called user-driven NMT. Specifically, a cache-based module and a user-driven contrastive learning method are proposed to offer NMT the ability to capture potential user traits from their histor- ical inputs under a zero-shot learning fashion. Furthermore, we contribute the first Chinese- English parallel corpus annotated with user behavior called UDT-Corpus. Experimental results confirm that the proposed user-driven NMT can generate user-specific translations. 1 1 Introduction In recent years, neural machine translation (NMT) models (Sutskever et al., 2014; Luong et al., 2015; Vaswani et al., 2017) have shown promising qual- ity and thus increasingly attracted users. When drawing on a translation system, every user has his own traits, including topic preference, stylistic characteristics, and expression habits, which can be implicitly embodied in their behavior, e.g., the historical inputs of these users. A good transla- tion should implicitly mirror user traits rather than ∗ Jinsong Su is the corresponding author. This work was done when Huan Lin was interning at DAMO Academy, Alibaba Group. mark1We release our source code and the associated bench- at https://github.com/DeepLearnXMU/ User-Driven-NMT. That is amazing ! Cool ! [cheerful, outgoing, active] [polite, formal, gentle] 太棒了！ 太棒了！ A B Figure 1: An example in which user traits leads to synonymous yet stylistically different translations. merely translate the original content, as the exam- ple shown in Figure 1. However, current NMT models are mainly designed for the semantic trans- formation between the source and target sentences regardless of subtle traits with respect to user be- havior. It can be said that the effect of user behavior on translation modeling is still far from utilization, which, to some extent, limits the applicability of NMT models in real-world scenarios. More recently, several studies have shown that the prominent signals in terms of personal char- acteristics can be served as inductive biases and reflected in translation results using domain adapta- tion approaches, such as personality (Mirkin et al., 2015), gender (Rabinovich et al., 2017), and po- liteness (Sennrich et al., 2016a). However, previ- …

Content truncated to 500 words