A model that uses BERT transformers and corresponding preprocessing layer that predicts whether a given tweet is "cancelling" someone (0), supporting/defending them (2) or is engaging with the cancelling in some way but is not cancelling or defending them, such as jokes, expressly non-committal statements, etc (1)
Also includes a text analysis component in R which can be used with the csv file included
The training dataset is based on the Lindsay Ellis Cancellation, with tweets manually annotated as critical, supportive or neutral.