This is modified version of UCBerkley DLab dataset, converted to represtent harmfull level in 3 classes:
SAFE - hate_speech_score < -1
BORDERLINE - -1 < hate_speech_score < 0.5
HARMFUL - hate_speech_score >= 0.5
Text and harmful level classification was created by UCBerkley DLab Team! I'm not creator of this data - i'm only converted it into classes!
