One Source, Two Targets: Challenges and Rewards of Dual Decoding

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 8533–8546 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 8533 One Source, Two Targets: Challenges and Rewards of Dual Decoding Jitao Xu Univ. Paris-Saclay, & CNRS, LISN Orsay, France jitao.xu@limsi.fr François Yvon Univ. Paris-Saclay, & CNRS, LISN Orsay, France francois.yvon@limsi.fr Abstract Machine translation is generally understood as generating one target text from an input source document. In this paper, we consider a stronger requirement: to jointly generate two texts so that each output side effectively de- pends on the other. As we discuss, such a device serves several practical purposes, from multi-target machine translation to the genera- tion of controlled variations of the target text. We present an analysis of possible implemen- tations of dual decoding, and experiment with four applications. Viewing the problem from multiple angles allows us to better highlight the challenges of dual decoding and to also thoroughly analyze the benefits of generating matched, rather than independent, translations. 1 Introduction Neural Machine Translation (NMT) is progress- ing at a rapid pace. Since the introduction of the first encoder-decoder architecture (Sutskever et al., 2014; Cho et al., 2014), then completed with an attention mechanism (Bahdanau et al., 2015; Vaswani et al., 2017), the performance of NMT systems is now good enough for a growing num- ber of services, both for the general public and the translation industry. Not only are neural transla- tion systems better, they are also more versatile and have been extended in many ways to meet new application demands. This is notably the case with multilingual extensions (Firat et al., 2016; Ha et al., 2016; Johnson et al., 2017), which aim to develop systems capable of processing multiple translation directions with one single model. Another common situation for MT applications is the multi-source / multi-target scenario, where source documents in language Sl need to be pub- lished in several target languages T 1 l , T 2 l , . . . . This is, for instance, what happens in multilingual insti- tutions, or with crowdsourced translations of TV shows. The multi-source way (Och and Ney, 2001; f I could do that again if you want . e1 只要 你 愿意 我 可以 重复 一遍 。 e2 もう 一 回 やり ましょ う か e1 Je peux le refaire si vous le voulez . e2 . voulez le vous si refaire le peux Je e1 Ich kann das noch mal machen , wenn Sie wollen . e2 Ich kann das noch mal machen , wenn du willst . Table 1: Instances of Dual Decoding: multi-target translation (§ 3), bi-directional decoding (§ 4), variant generation (§ 6). Schwartz, 2008; Zoph and Knight, 2016) to handle this generates a first translation into target language T 1 l , which, once revised, can be used in conjunction with the original source to generate the translation into language T 2 l . The expected benefit of this approach is to facilitate word disambiguation. An alternative, that …

Content truncated to 500 words