HateBERT
BERT retrained on banned Reddit communities for abusive language detection.
About HateBERT
The HateBERT repository on GitHub is listed under the account tommasoc80. This GitHub repository is a mirror for HateBERT.
The retrained model, the fine-tuned models, data, and code are available on an Open Science Fundation repository. The page also links to a paper about the work.
Drafted from github.com · not yet confirmed by HateBERT
How it works
HateBERT is a BERT model further pre trained on posts from banned Reddit communities, then fine tuned for offensive, abusive and hateful language detection. It's a research baseline that beats plain BERT on the standard hate speech datasets, and it's small enough to run anywhere. English only.
Reviews
Be the first to review it
AnonymousKnow a team weighing HateBERT? Send them this page.