Natural Language Processing TODO - A LOT to add! Holy $#!t: Are popular toxicity models simply profanity detectors?