ETHOS
multi-labEl haTe speecH detectiOn dataSet
TextsGNU GPLv3
ETHOS is a hate speech detection dataset. It is built from YouTube and Reddit comments validated through a crowdsourcing platform. It has two subsets, one for binary classification and the other for multi-label classification. The former contains 998 comments, while the latter contains fine-grained hate-speech annotations for 433 comments.
Source: https://arxiv.org/abs/2006.08328
Related Benchmarks
Ethos Binary/Abuse Detection/Classification AccuracyEthos Binary/Abuse Detection/F1-scoreEthos Binary/Abuse Detection/PrecisionEthos Binary/Hate Speech Detection/Classification AccuracyEthos Binary/Hate Speech Detection/F1-scoreEthos Binary/Hate Speech Detection/PrecisionEthos MultiLabel/Abuse Detection/Hamming LossEthos MultiLabel/Hate Speech Detection/Hamming Loss