Working with Pre-translated Texts: Preliminary Findings from a Survey on Post-editing and Revision Practices in Swiss Corporate In-house Language Services

Working with Pre-translated Texts: Preliminary Findings from a Survey on Post-editing and Revision Practices in Swiss Corporate In-house Language Services Sabrina Girletti FTI/TIM, University of Geneva Switzerland sabrina.girletti@unige.ch Abstract With the arrival of neural machine transla- tion, the boundaries between revision and post-editing (PE) have started to blur (Ko- ponen et al., 2020). To shed light on current professional practices and provide new pedagogical perspectives, we set up a survey-based study to investigate how PE and revision are carried out in profes- sional settings. We received 86 responses from corporate translators working at 23 different corporate in-house language ser- vices in Switzerland. Although the differ- ences between the two activities seem to be clear for in-house linguists, our findings show that they tend to use the same read- ing strategies when working with human- translated and machine-translated texts. 1 Introduction In recent years, quality improvements achieved by the latest-generation machine translation systems have put machine translation (MT) under the spot- light. Results of recent language industry sur- veys (ELIS, 2022; Pielmeier and Lommel, 2019) show that language service providers (LSPs) iden- tify MT post-editing (PE) as one of the most re- quested services and as an opportunity to increase productivity and improve profit margins. There- fore, many of them have implemented MT or plan to do so as soon as possible. In Switzerland, a multilingual country where many companies have their own in-house transla- tion service, the situation is no different. Many © 2022 The authors. This article is licensed under a Creative Commons 3.0 licence, no derivative works, attribution, CC- BY-ND. LSPs have already added MT to their workflows and started offering PE among their services, to- gether with translation and revision. Since neural MT (NMT) output more closely resembles human translations than machine- translated texts (Martikainen, 2019; Yamada, 2019), correcting it is often considered more sim- ilar to a revision. Recent work by Koponen et al. (2020) has paved the way for studying the relation- ship between these two activities whose bound- aries are “starting to blur” (2020:3). To shed light on current practices and pro- vide new perspectives for the training of both stu- dents and experienced translators who work with MT, we set up a survey-based study to investi- gate how PE and revision are carried out in pro- fessional settings. In particular, we chose to focus on Switzerland-based corporate in-house language services (CILS), as this cohort is underrepresented in language industry surveys and has been scarcely investigated compared to institutional (Cadwell et al., 2017; Riondel, 2021; Rossi and Chevrot, 2019) and freelance translators (Gaspari et al., 2015; Zaretskaya, 2015). Our study consisted of two questionnaires, available in four languages: the first questionnaire (Q1) was aimed at language service directors and project managers and contained questions about the structure and workflow of the language service. The second questionnaire (Q2) was aimed at lan- guage service employees who translate, revise and post-edit texts. It included questions about their workflows, strategies and attitudes towards PE and revision. In …

Content truncated to 500 words