I think this and Twitter's attempts to police the language of its users is quite flawed since context is hard to read into a sentence if you're doing the most naive analysis as it seems both are doing right now. I doubt throwing ML or NLP at this problem will ever work because the biggest problem with moderation is that people tend to get into fights without much intention of getting into one. So for me moderation is a human-level concern that should be solely done by humans since human moderators despite their own biases are better judges of misdeeds and outright trolling. Once FB and Twitter get on board with this idea and stop pretending it's a technology problem the sooner their parts of the Web might be tolerable.