POLAR (Polemics in Online Language and Automated Recognition) is a 24-month collaborative research project funded within the YUFE Alliance and hosted at Université Sorbonne Nouvelle.
The project investigates how online speakers express stance and controversy through covert or overt discursive strategies (e.g., distancing, alignment, rhetorical questions or remarks), and how such strategies can be identified qualitatively and then detected automatically.
Research aims
Distance and controversy markers functioning as polemic markers, and model how they operate in context.
Tool-assisted approaches — automatic annotation, detection, clustering and text mining — to refine automatic decision-making for polemic detection.
Method in a nutshell
Manual identification of polemical strategies across linguistic, abstract (e.g., irony/metaphor), and argumentation levels.
Annotation, machine learning / NLP, and evaluation on large-scale social media data.
Planned activities
A series of interdisciplinary webinars (4–5 across the project)
Regular collaboration meetings and in-person workshops
Final colloquium / conference to present results and foster future collaborations