Detects hate speech (attacks on people because of race or ethnicity, origin, religion, gender, sexual orientation, disability or age) and the targeted group, in English and about 50 other languages, with multilingual-hate-speech-base (308M, Apache-2.0), running entirely in your browser. Insults, profanity and counter-speech are not hate speech. One text per line; scores at or above 0.5 are bold.
Loading model…
What counts as hate speech depends on law, policy and culture; the model can miss implicit hate and misjudge quotes. Target scores are meaningful only when the hate score is high. See the model card for measured accuracy per language.