LIG-CRIStAL Submission for the WMT 2017 Automatic Post-Editing Task

Proceedings of the Conference on Machine Translation (WMT), Volume 2: Shared Task Papers, pages 623–629 Copenhagen, Denmark, September 711, 2017. c⃝2017 Association for Computational Linguistics LIG-CRIStAL System for the WMT17 Automatic Post-Editing Task Alexandre B´erard Olivier Pietquin Univ. Lille, CNRS, Centrale Lille, Inria, UMR 9189 CRIStAL alexandre.berard@ed.univ-lille1.fr olivier.pietquin@univ-lille1.fr Laurent Besacier LIG, Univ. Grenoble Alpes, CNRS laurent.besacier@univ-grenoble-alpes.fr Abstract This paper presents the LIG-CRIStAL submission to the shared Automatic Post- Editing task of WMT 2017. We propose two neural post-editing models: a mono- source model with a task-specific atten- tion mechanism, which performs particu- larly well in a low-resource scenario; and a chained architecture which makes use of the source sentence to provide extra con- text. This latter architecture manages to slightly improve our results when more training data is available. We present and discuss our results on two datasets (en-de and de-en) that are made available for the task. 1 Introduction It has become quite common for human translators to use machine translation (MT) as a first step, and then to manually post-edit the translation hypoth- esis. This can result in a significant gain of time, compared to translating from scratch (Green et al., 2013). Such translation workflows can result in the production of new training data, that may be re-injected into the system in order to improve it. Common ways to do so are retraining, incremental training, translation memories, or automatic post- editing (Chatterjee et al., 2015). In Automatic Post-Editing (APE), the MT sys- tem is usually considered as a blackbox: a separate APE system takes as input the outputs of this MT system, and tries to improve them. Statistical Post- Editing (SPE) was first proposed by Simard et al. (2007). It consists in training a Statistical Machine Translation (SMT) system (Koehn et al., 2007), to translate from translation hypotheses to a human post-edited version of those. B´echara et al. (2011) then proposed a way to integrate both the transla- tion hypothesis and the original (source language) sentence. More recent contributions in the same vein are (Chatterjee et al., 2016; Pal et al., 2016). When too little training data is available, one may resort to using synthetic corpora: with sim- ulated PE (Potet et al., 2012), or round-trip translation (Junczys-Dowmunt and Grundkiewicz, 2016). Recently, with the success of Neural Machine Translation (NMT) models (Sutskever et al., 2014; Bahdanau et al., 2015), new kinds of APE methods have been proposed that use encoder-decoder ap- proaches (Junczys-Dowmunt and Grundkiewicz, 2016, 2017; Libovick´y et al., 2016; Pal et al., 2017; Hokamp, 2017), in which a Recurrent Neu- ral Network (RNN) encodes the source sequence into a fixed size representation (encoder), and an- other RNN uses this representation to output a new sequence. These encoder-decoder models are generally enhanced with an attention mechanism, which learns to look at the entire sequence of en- coder states (Bahdanau et al., 2015; Luong et al., 2016). We present novel neural architectures for au- tomatic post-editing. Our models learn to gener- ate sequences of edit operations, and use …

Content truncated to 500 words