ArgMine

NLP: Argumentation mining & propaganda detection

There’s a lot of misleading content around, and identifying it automatically requires solving some big NLP problems: how can we extract and thus automatically evaluate argumentation? What does misleading or malicious argumentation look like from a computational point of view? Identifying manipulative content is possible for humans, but this does not scale to the web; meanwhile, the web is rife with false and deceptive claims. Computational approaches offer the ability to scale up the search for this content without needing extra human effort.
This project investigates basic general methods and models for extracting and modelling argumentation from documents and social media conversations, and for separating good- and poor-faith arguments based on these representations.