On the definition and the automatic detection of hate speech. A tale of pulling translators into NLP

The democratisation of the web has opened the door for people to freely express online. Unfortunately, malicious users take advantage of it to spread hate. In this talk, I will present our efforts on the creation of automatic approaches for the automatic identification of different kinds of hate speech over the last three years. I will pay special attention to discuss open questions, such as where is the boundary between hate speech and “freedom of expression”, which actions should be taken when in the presence of a hateful post or community, and what exactly means hate speech (but from a societal and from a legal perspective). Aspects that make deep learning models struggle to properly identify hate speech (such as lexical creativity, lack of annotated data or level of expliciteness).

Nov 17, 2023 4:40 PM — 5:00 PM