Evaluating Gender Bias in German Machine Translation
Michelle Kappl, Technische Universität Berlin, Softwaretechnik und Theoretische Informatik
Inhalte
Every day, millions rely on machine translation (MT) models to overcome language barriers, but can we trust them to be fair, or do they perpetuate harmful stereotypes? The goal of this project is to refine an evaluation method for assessing gender bias, specifically underrepresentation and stereotyping, in German MT. Central to this is a dataset that we will expand and improve, e.g. by integrating gender-neutral language plus better cultural and linguistic accuracy. Furthermore, we revise the automatic evaluation process of MT systems. We will then test popular MT models such as DeepL, and LLMs like ChatGPT for potential biases. Mid-Bachelor or Master students in computer science, linguistics, or social sciences with an interest in MT, gender studies, and fair technology are invited to join this project. Basic Python skills are a plus, but a willingness to learn is the most important thing. Let’s contribute to the development of fairer, more inclusive translation technologies together!
Fachliche:r Betreuer:in
Dr.-Ing. Stefan Hillmann
Kontakt
michelle.kappl@tu-berlin.de