Despite the increasingly good quality of automatic translations, machine-translated texts require corrections. Automatic post-editing models have been introduced to perform these corrections without human intervention. However, no system has been able to fully automate the post-editing process. Moreover, while numerous translation tools benefit from translators’ input, human-computer interaction has been underexplored in post-editing.
This talk will discuss automatic post-editing models and suggest that they could be improved in more interactive scenarios, as previously done in machine translation. While some attempts were made to update automatic post-editing models incrementally, this was mostly done using synthetic corpora, which is likely to affect the performance. To address this issue and as part of this project, automatic post-editing models trained in a traditional setting were developed and updated in both batch and online modes without using artificial resources, with a view to analysing the performance of incremental adaptations in different systems, domains and language pairs. While the interaction with the translator was simulated, an interactive functionality allowing for dynamic post-editing was included for demonstration purposes. The results showed that none of the models was able to beat the baseline and that the online models systematically yielded a lower performance. Moreover, this study provided a human evaluation of the outputs obtained in both batch and online models, which constitutes a significant contribution, given that online models tend to be examined using automatic metrics only. This evaluation allowed for identifying recurrent error patterns, such as incorrect deletions, insertions and substitutions, as well as errors related to sentence structure and figurative language.
Such outcomes confirm the difficulties faced in automatic post-editing. Based on the results, several recommendations will be put forward for conducting further research, including experiments with more data and different environmental variables.
Marie Escribe holds a BA in Applied Foreign Languages from the University of Haute-Alsace and an MA in Translation from London Metropolitan University. She has worked as a freelance translator specialised in scientific and technical fields for more than two years. Since 2019, she has indeed established strong collaborations with several LSPs and has been entrusted with various responsibilities, going from translation and transcreation to post-editing and project management. Her experience working with machine translation and CAT tools combined with her strong interest in language technologies led her to another Master’s in Computational Linguistics at the Research Group in Computational Linguistics, University of Wolverhampton, which she completed in 2021. Her research interests revolve around translation technologies and include in particular post-editing, translation memory systems and translation quality evaluation.