Entretien
SPOTLIGHT ON鈥� PERRINE SCHUMACHER
Perrine Schumacher graduated from the AV短视频 of Liège, Belgium, with a degree in translation. Her working languages are French, English, and German. She completed her doctorate jointly at Liège in languages, literature and translation studies and at UNIGE in translation technology. Her main research interests are emergent translation technologies and translation didactics.
Her dissertation, “La post-édition de traduction automatique en contexte d’apprentissage : Effets sur la qualité et défis pour l’enseignement de la traduction “ [Post-editing machine translation in learning settings: Impacts on quality and challenges for teaching translation], explores the impact of post-editing on translation quality and the challenges it poses for teaching translation. Perrine was recently awarded the Latsis Prize, awarded annually by Swiss universities for outstanding scholarship. She spoke to us about her dissertation and her research projects.
Lisez cet entretien en français
You recently won the Latsis Prize for 2024. Can you tell us a bit more about your journey to that point?
I began my studies with a BA and MA in translation at the AV短视频 of Liège, followed by two years working as a private sector translator. From 2016 to 2023 I held an assistantship while working on my PhD at Liège, writing the dissertation on post-editing machine translation that won me the prize. Post-editing can be defined as human intervention to modify and improve automatically generated machine translation. The research gave me a specialisation in a fast-growing area of study and let me add to the body of thought on modernising translation teaching.
What led you to a PhD partnership?
Back in 2016, when I told my research supervisor, Prof. Valérie Bada, that I wanted to work on neural machine translation, I quickly realised that it was a complex, fast-moving area of research that called for joint supervision. That was the starting point for my scholarly venture into pastures new. At that point, I could not find a specialist in emergent translation technologies in Wallonia. I started looking further afield and I found what I needed in Switzerland. Partnering with the FTI let me tap into the expertise of my Swiss colleagues and especially the valuable support and advice of my co-supervisor, Prof. Pierrette Bouillon. We first met in 2017 at the Solbosch campus at the Université libre de Bruxelles (ULB), where I outlined my project to her. And I have to admit, MT and I have been together ever since!
Your research is right up to date, especially with the rise of NMT and tools based on generative AI. Can you tell us briefly why your research is significant to your field?
I think my work has sparked some interest because it raises the important question of where humans fit into a digital sector undergoing rapid change and a world of ever faster paced technology. I wanted to foreground the need to bring translation training courses up to date to familiarise translators with changing technologies and with the many issues at stake in emergent technologies. I also hope my research will help spotlight the added value of translators compared to machines that translate automatically. I would also like to point out that currently, no technology can produce publication-quality translations without human input. Humans are still at the heart of the translation process and must remain there. It’s crucial to be aware of that and to pass that knowledge on.
What are the most striking findings from your research?
Definitely the discovery of a levelling effect on quality in post-editing. I became aware of an inverse relationship between the level of translation students and the quality of their post-editing work: the weaker their translation skills, the more they made use of post-editing, and conversely, the broader their translation skills, the more post-editing lowered the quality of the final product.
My results also demonstrated the existence of “post-editese” in the corpus – linguistic traits typical of post-edited language, distinct from translated language and raw native language (French, in my case). Compared to human translations, post-edited NMT (in particular texts translated by DeepL) proved less lexically rich and syntactically closer to source texts (English in my research).
These observations led me to put forward a considered approach to MT tools, helping translation students develop tech knowledge and skills valued on the job market. I also outlined the pros and cons of AI technology, transcending the binary opposition between human and machine shared by the media and tech businesses and building capacity for adaptability, whatever the tool.
The Latsis Prize is wonderful recognition of your work. What is next for you, after this award?
The Latsis Prize is an amazing honour that will really boost my future research. The prize and the F.R.S-FNRS (Belgium) postdoctoral researcher position I have just started mean I can develop a new project bringing in generative AI. I’d like to explore the potential of chatbots like ChatGPT and Gemini as new tools for translation and post-editing support. These tools, building on generative AI, raise issues and challenges on the same lines as NMT, once again asking questions about the future of professional translation and translation training. Such powerful chatbots are freely available to all users, enabling them to combine MT and post-editing and style improvement that keeps getting better, so I think it is urgent to study their impact on the quality of the finished product and on the process of translation itself, and to encourage using them in a considered, responsible manner.
For now, I am determined to keep demystifying AI for upcoming generations of translators who, as we know, are particularly attuned to the dominant marketing discourse that tends to embellish reality and overestimate the performance of these new technologies. It’s therefore essential to show them that humans still have a central, crucial role to play, particularly in translation. The aim is to train a new generation of responsible translators trained in critical thinking, able to use new technologies to their best advantage while making the most of the unique human input that remains at the heart of the profession.