Scores comments for toxicity, severe_toxicity, obscene, threat, insult, identity_attack, sexual_explicit (the Detoxify labels) in English and 33 other languages with multilingual-toxicity-small (141M, Apache-2.0), running entirely in your browser. One comment per line; labels at or above 0.5 are bold.
Loading model…
Toxicity judgements are subjective; the model can over-flag mentions of identity groups and miss sarcasm. See the model card for measured accuracy per language.