Matthieu Labeau - Representing and processing rare events in NLP models - a few research directions
- Date: 14 avril 2025 à 13h
- Salle: 65-66 304
My research explores how to propose efficient NLP methods in settings where data is limited and annotations are sparse. I focus on developing frugal, low-supervision methods for learning under resource constraints: I will first present tailored strategies for imbalanced classification, and follow with preliminary results on anomaly detection on text. In parallel, I am interested in interactional data, and will quickly present work on improving dialogue agents strategy prediction through modular approaches.