Findings of the WMT 2020 Biomedical Translation Shared Task: Basque, Italian and Russian as New Additional Languages

Proceedings of the 5th Conference on Machine Translation (WMT), pages 660–687 Online, November 19–20, 2020. c⃝2020 Association for Computational Linguistics 660 Findings of the WMT 2020 Biomedical Translation Shared Task: Basque, Italian and Russian as New Additional Languages Rachel Bawden1∗ Giorgio Maria Di Nunzio2 Cristian Grozea3 Iñigo Jauregi Unanue4 Antonio Jimeno Yepes5 Nancy Mah6 David Martinez5 Aurélie Névéol7 Mariana Neves8,9 Maite Oronoz10 Olatz Perez de Viñaspre10 Massimo Piccardi4 Roland Roller11 Amy Siu12 Philippe Thomas11 Federica Vezzani13 Maika Vicente Navarro14 Dina Wiemann15 Lana Yeganova16 1School of Informatics, University of Edinburgh, Scotland 2Dept. of Information Engineering, University of Padua, Italy 3Fraunhofer Institute FOKUS, Berlin, Germany 4University of Technology Sydney, Sydney, Australia 5IBM Research Australia, Melbourne, Australia 6Fraunhofer Institute for Biomedical Engineering (IBMT), Berlin, Germany 7LIMSI, CNRS, Université Paris-Saclay, Orsay, France 8German Centre for the Protection of Laboratory Animals (Bf3R), 9German Federal Institute for Risk Assessment (BfR), Berlin, Germany 10IXA NLP Group, University of the Basque Country, Donostia, Spain 11German Research Center for Artificial Intelligence (DFKI), Berlin, Germany 12Beuth University of Applied Sciences, Berlin, Germany 13Dept. of Linguistic and Literary Studies University of Padua, Italy 14Maika Spanish Translator, Melbourne, Australia 15Novartis AG, Basel, Switzerland 16NCBI/NLM/NIH, Bethesda, USA Abstract Machine translation of scientific abstracts and terminologies has the potential to sup- port health professionals and biomedical researchers in some of their activities. In the fifth edition of the WMT Biomedical Task, we addressed a total of eight language pairs. Five language pairs were previously addressed in past editions of the shared task, namely, English/German, English/French, English/Spanish, English/Portuguese, and English/Chinese. Three additional lan- guages pairs were also introduced this year: English/Russian, English/Italian, and English/Basque. The task addressed the evalu- ation of both scientific abstracts (all language pairs) and terminologies (English/Basque only). We received submissions from a total of 20 teams. For recurring language pairs, we observed an improvement in the translations in terms of automatic scores and qualitative evaluations, compared to previous years. ∗ The author list is alphabetical and does not reflect the respective author contributions. 1 Introduction Automatic translation aims to alleviate the lan- guage barrier by providing access to information for readers not familiar with the original language used to write documents. Access to accurate biomedical information is specifically critical and machine translation (MT) can contribute to making health information available to health profession- als and the general public in their own language. It can also contribute to biomedical research by assisting with the writing of research reports in English. In addition, machine translation can pro- vide the opportunity to enhance the use of natu- ral language processing (NLP) tools and methods for low-resource languages by the development of resources through translation or by making tools available through text translation into resource rich languages. Herein, we describe the fifth edition of the WMT Biomedical task,1 which aims to evaluate the auto- 1http://www.statmt.org/wmt20/ 661 matic translation of a variety of biomedical texts. The first edition of the task (Bojar et al., 2016) focused on biomedical scientific abstracts in three language pairs. The second …

Content truncated to 500 words