Findings of the WMT 2022 Biomedical Translation Shared Task: Monolingual Clinical Case Reports

Proceedings of the Seventh Conference on Machine Translation (WMT), pages 694 - 723 December 7-8, 2022 ©2022 Association for Computational Linguistics Findings of the WMT 2022 Biomedical Translation Shared Task: Monolingual Clinical Case Reports Mariana Neves1 ∗Antonio Jimeno Yepes2 Amy Siu3 Roland Roller4 Philippe Thomas4 Maika Vicente Navarro5 Lana Yeganova6 Dina Wiemann7 Giorgio Maria Di Nunzio8 Federica Vezzani8 Christel Gerardin9 Rachel Bawden10 Darryl Johan Estrada11 Salvador Lima-López11 Eulàlia Farré-Maduell11 Martin Krallinger11 Cristian Grozea12 Aurélie Névéol13 1German Centre for the Protection of Laboratory Animals (Bf3R), German Federal Institute for Risk Assessment (BfR), Berlin, Germany 2RMIT University, Australia 3Berliner Hochschule für Technik, Germany 4German Research Center for Artificial Intelligence (DFKI), Berlin, Germany 5Leica Biosystems, Australia 6NCBI/NLM/NIH, Bethesda, USA 7Novartis AG, Basel, Switzerland 8Dept. of Linguistic and Literary Studies University of Padua, Italy 9Sorbonne Université, Inserm, IPLESP, Paris, France 10Inria, Paris, France 11Barcelona Supercomputing Center, Spain 12Fraunhofer Institute FOKUS, Berlin, Germany 13Université Paris-Saclay, CNRS, LISN, Orsay, France Abstract In the seventh edition of the WMT Biomedical Task, we addressed a total of seven language pairs, namely English/German, English/French, English/Spanish, English/Portuguese, En- glish/Chinese, English/Russian, English/Italian. This year’s test sets covered three types of biomedical text genre. In addition to scientific abstracts and terminology items used in previ- ous editions, we released test sets of clinical cases. The evaluation of clinical cases transla- tions were given special attention by involving clinicians in the preparation of reference trans- lations and manual evaluation. For the main MEDLINE test sets, we received a total of 609 submissions from 37 teams. For the ClinSpEn sub-task, we had the participation of five teams. ∗The contribution of the authors are the following: MN prepared the MEDLINE test sets, performed test set validation, manual validation, and organized the task; AJY performed test sets validation, manual validation and the automatic eval- uation; RR, PT, MVN, LY, DW. GMDN, FV performed test sets validation and manual validation; CGn created reference translation and performed manual validation; RB performed manual validation; DJE, SLL, EFM, MK organized the Clin- SpEn sub-task; CGa created the baselines; and AN collected information on participants’ methods, performed test sets val- idation, manual validation and created reference translation. All authors approved the final version of the manuscript. E- mail for contact: mariana.lara-neves@bfr.bund.de 1 Introduction This is the seventh edition of the biomedical trans- lation task offered under the umbrella of the Con- ference on Machine Translation (WMT22).1 This shared task builds on the six previous editions of the biomedical translation task (Bojar et al., 2016; Jimeno Yepes et al., 2017; Neves et al., 2018; Baw- den et al., 2019, 2020; Yeganova et al., 2021). Sim- ilar to previous years, we addressed seven language pairs, in both directions, namely: German/English (de2en and en2de), Spanish/English (es2en and en2es), French/English (fr2en and en2fr), Ital- ian/English (it2en and en2it), Portuguese/English (pt2en and en2pt), Russian/English (ru2en and en2ru), and Chinese/English (zh2en and en2zh). In the biomedical translation task this year, par- ticipants were asked to translate shared test sets (described in Section 2) comprising documents be- longing …

Content truncated to 500 words