Contextual Neural Model for Translating Bilingual Multi-Speaker Conversations

Proceedings of the Third Conference on Machine Translation (WMT), Volume 1: Research Papers, pages 101–112 Belgium, Brussels, October 31 - Novermber 1, 2018. c⃝2018 Association for Computational Linguistics https://doi.org/10.18653/v1/W18-64011 Contextual Neural Model for Translating Bilingual Multi-Speaker Conversations Sameen Maruf1 Monash University VIC, Australia Andr´e F. T. Martins2 Unbabel & Instituto de Telecomunicac˜oes Lisbon, Portugal 1{firstname.lastname}@monash.edu 2andre.martins@unbabel.com Gholamreza Haffari1 Monash University VIC, Australia Abstract Recent works in neural machine transla- tion have begun to explore document trans- lation. However, translating online multi- speaker conversations is still an open problem. In this work, we propose the task of trans- lating Bilingual Multi-Speaker Conversations, and explore neural architectures which exploit both source and target-side conversation histo- ries for this task. To initiate an evaluation for this task, we introduce datasets extracted from Europarl v7 and OpenSubtitles2016. Our ex- periments on four language-pairs confirm the significance of leveraging conversation his- tory, both in terms of BLEU and manual eval- uation. 1 Introduction Translating a conversation online is ubiquitous in real life, e.g. in the European Parliament, United Nations, and customer service chats. This sce- nario involves leveraging the conversation history in multiple languages. The goal of this paper is to propose and explore a simplified version of such a setting, referred to as Bilingual Multi-Speaker Machine Translation (Bi-MSMT), where speak- ers’ turns in the conversation switch the source and target languages. We investigate neural architec- tures that exploit the bilingual conversation history for this scenario, which is a challenging problem as the history consists of utterances in both lan- guages. The ultimate aim of all machine translation systems for dialogue is to enable a multi-lingual conversation between multiple speakers. How- ever, translation of such conversations is not well- explored in the literature. Recently, there has been work focusing on using the discourse or docu- ment context to improve NMT, in an online set- ting, by using the past context (Jean et al., 2017; Wang et al., 2017; Bawden et al., 2017; Voita et al., 2018), and in an offline setting, using the past and future context (Maruf and Haffari, 2018). In this paper, we design and evaluate a conversa- tional Bi-MSMT model, where we incorporate the source and target-side conversation histories into a sentence-based attentional model (Bahdanau et al., 2015). Here, the source history comprises of sen- tences in the original language for both languages, and the target history consists of their correspond- ing translations. We experiment with different ways of computing the source context represen- tation for this task. Furthermore, we present an effective approach to leverage the target-side con- text, and also present an intuitive approach for incorporating both contexts simultaneously. To evaluate this task, we introduce datasets extracted from Europarl v7 and OpenSubtitles2016, con- taining speaker information. Our experiments on English-French, English-Estonian, English- German and English-Russian language-pairs show improvements of +1.44, +1.16, +1.75 and +0.30 BLEU, respectively, for our best model over the context-free baseline. The results show the im- pact of conversation history on translation of bilin- gual multi-speaker conversations …

Content truncated to 500 words