CUNI Systems for WMT21: Terminology Translation Shared Task

Proceedings of the Sixth Conference on Machine Translation (WMT), pages 828–834 November 10–11, 2021. ©2021 Association for Computational Linguistics 828 CUNI systems for WMT21: Terminology translation Shared Task Josef Jon and Michal Novák and João Paulo Aires and Dušan Variš and Ondˇrej Bojar Charles University {jon,aires,varis,bojar}@ufal.mff.cuni.cz Abstract This paper describes Charles University sub- mission for Terminology translation Shared Task at WMT21. The objective of this task is to design a system which translates cer- tain terms based on a provided terminology database, while preserving high overall trans- lation quality. We competed in English-French language pair. Our approach is based on pro- viding the desired translations alongside the input sentence and training the model to use these provided terms. We lemmatize the terms both during the training and inference, to al- low the model to learn how to produce correct surface forms of the words, when they differ from the forms provided in the terminology database. Our submission ranked second in Exact Match metric which evaluates the abil- ity of the model to produce desired terms in the translation. 1 Introduction Terminology integration, or, more generally, con- strained translation in NMT was extensively stud- ied in recent years. Lexically constrained transla- tion means that aside from the source sentence, we have available some additional knowledge of what tokens or expressions should appear in the trans- lation and we want to force the system to include them in the generated output. Three main ways of enforcing these constraints have been studied. First, replacing the source part of the constraint that is found in the source sentence with a place- holder which is then copied by the model into the output. There it gets replaced by the target part of the constraint (Luong et al. (2015); Crego et al. (2016)). Second way is to modify the decoding search algorithm in a way that only allows hypotheses containing the constraints to be marked as finished (Anderson et al. (2017); Hasler et al. (2018); Chat- terjee et al. (2017); Hokamp and Liu (2017); Post and Vilar (2018); Hu et al. (2019)) Finally, some works focus on providing the con- straints directly to the model as part of the input sequence. The model is trained to incorporate these constraints into the output, for example Dinu et al. (2019); Chen et al. (2020); Song et al. (2019) or Bergmanis and Pinnis (2021). As apparent from previous paragraphs, the prob- lem of integrating lexical constraints into NMT is well studied, but one issue was largely ignored. In inflected languages, the surface form of the con- straint in the output cannot be known beforehand, as there are usually many possible ways to trans- late a sentence and many of them need different surface forms of the constraint to be fluent and grammatically correct. For example, let’s say we have a terminology database containing term pair influenza -> grippe and this source sentence: During the 2018-2019 influenza season. Possible correct translation is: Pendant la saison grippale 2018-2019. Where the term …

Content truncated to 500 words