The Ergonomics of Literary Machine Translation: Effects of Technology on Translator Flow

Since the advent of neural machine translation (NMT), more and more research has delved into the topic of literary machine translation, focusing more recently on the possibility of customizing these tools for this particular activity (Kenny and Winters 2024). In this context, our previous work aimed to assess the technical feasibility of individualized and style-adaptive NMT. While the conclusions of this work were largely positive, further experiments were recently carried out to reinforce the ecological validity of the research, by conducting an expert evaluation during which we asked a professional literary translator to translate three chapters with the help of this customized system. To that end, we adopted the methodology presented by Borg (2023), combining various investigation methods to uncover the cognitive and creative processes of a translator working in situ.

These observation and interview sessions revealed many constraints influencing the translation, and were further confirmed by the comparison of the same excerpt translated without the help of MT ten years ago. Unsurprisingly, this expert evaluation highlighted and reactualized the exact same types of cognitive loads, both intrinsic and extraneous, that were already identified with the growing use of CAT tools (O'Brien et al. 2017), and that are mostly tied to the post-editing interface, on the one hand, and the modalities of the human-machine interaction on the other. These results thus highlight the dire need for more user-centred research, especially considering how little innovation has to do with considerations such as ease of use or work satisfaction (Lavault-Olléon 2011), as well as more work involving both MT and CAT tools (Vieira et al. 2023), focusing on the interfaces and interactions of machine translation systems and the post-editing activity rather than purely technical improvements (Laubli et al. 2022), and promoting more cooperation between researchers, developers and translators (O'Brien and Conlan 2018).