LMU Munich’s Neural Machine Translation Systems for News Articles and Health Information Texts

Proceedings of the Conference on Machine Translation (WMT), Volume 2: Shared Task Papers, pages 315–322 Copenhagen, Denmark, September 711, 2017. c⃝2017 Association for Computational Linguistics LMU Munich’s Neural Machine Translation Systems for News Articles and Health Information Texts Matthias Huck, Fabienne Braune, Alexander Fraser Center for Information and Language Processing LMU Munich Munich, Germany {mhuck,braune,fraser}@cis.lmu.de Abstract This paper describes the LMU Mu- nich English→German machine transla- tion systems. We participated with neural translation engines in the WMT17 shared task on machine translation of news, as well as in the biomedical translation task. LMU Munich’s systems deliver compet- itive machine translation quality on both news articles and health information texts. 1 Introduction The Center for Information and Language Pro- cessing at LMU Munich has a strong track record at building statistical machine translation (SMT) systems for various language pairs, e.g. for trans- lation between English and German, Czech, Ro- manian, Russian, or French. LMU has fre- quently participated in WMT machine translation shared tasks in recent years (Bojar et al., 2016, 2015, 2014, 2013), competing (and also collabo- rating) internationally in an open evaluation cam- paign with other leading research labs from both academia and industry. Research on various different types of machine translation models has previously been conducted at LMU. Core SMT paradigms for LMU’s past shared task participations include phrase-based models (Cap et al., 2015, 2014b; Weller et al., 2013; Sajjad et al., 2013), hierarchical phrase- based models (Huck et al., 2016; Peter et al., 2016), operation sequence models (Durrani et al., 2013), and hybrids of statistical approaches with rule-based and deep syntactic components (Tam- chyna et al., 2016b). At this year’s EMNLP 2017 Second Confer- ence on Machine Translation (WMT17),1 LMU participated in two shared tasks: the shared task 1http://www.statmt.org/wmt17/ on machine translation of news and the biomed- ical translation task. We submitted the output of our English→German machine translation sys- tems. The system for the news task was trained under “constrained” conditions, employing only permissible resources as defined by the shared task organizers. The system for the biomedical task builds upon our news task system, but was domain-adapted towards the medical domain via the usage of additional parallel training data from the in-domain sections of the UFAL Medical Cor- pus v.1.0. We have trained neural machine translation (NMT) models this year. Neural network mod- els for machine translation (Sutskever et al., 2014; Bahdanau et al., 2014) are now largely successful for many language pairs and domains. This has for instance become apparent with the University of Edinburgh’s excellent results in the WMT16 news translation shared task with neural systems (Sennrich et al., 2016a), which outperformed most other submitted systems, including Edinburgh’s own traditional SMT engines (Williams et al., 2016). LMU’s English→German neural machine translation systems confirm this trend. We have achieved competitive performance—in terms of translation quality as measured with BLEU (Pap- ineni et al., 2002)—in both shared tasks that we participated in.2 A unique characteristic of the LMU English→German NMT systems is a linguis- tically informed, …

Content truncated to 500 words