Hate Speech

Use argumentative reasoning to make the implicit content of hate speech explicit.

Measure how different definitions of abusive and hate speech affect detection models.

GenSPP trains the generator and predictor of a self-explaining text classifier separately with genetic search, removing interlocking and producing more accurate highlights.

HateDefCon breaks 493 hate speech definitions from five domains into semantic components and shows LLM hate speech detection changes with the definition in the prompt.

PejorativITy labels Italian tweets for pejorative words and misogyny, and shows that telling insults from neutral senses first improves misogyny detection.