Predicting hate and rationales via the language model BLOOM

This notebook shows how BLOOM can be used to detect hateful content (some examples taken from HateXplain) via the Inference API. Thanks to Sentdex for showing how to call BLOOM. Models are also available as downloads (check the size!). The model can also be tried out interactively. For more technical information about the model, check out this blogpost.

I am interested in NLP and working on my Master Thesis (Explainability in Hate Speech Detection) at TU Vienna. There are several different projects I contribute to (HateFlux, tuw_nlp, POTATO).

I am also creating desktop and mobile games, check out my website seismic.at!