A T5ForConditionalGeneration trained on 2 tasks from PAN Profiling Hate Speech Spreaders on Twitter dataset (EN):

  • topic attribution - topics were assigned with BertTopic library using embeddings from cardiffnlp/bertweet-base-hate Roberta model (train and test sets from the PAN task)
  • hate speech identification (train set from the PAN task)

in order to generate tone of comment use prefix hater classification:

Downloads last month
Hosted inference API
Text2Text Generation
This model can be loaded on the Inference API on-demand.