We could just filter that stuff the same way we filter email. Naive Bayes works at email scale why not comment scale? It should all be done on a per user basis so the filters only filter out language you personally find offensive.

It never made sense why try to force me to use the same filter as the Karen that complains about kids skateboarding.