Hate Speech may be a "Narrow World"

I hope to study hate speech and build a Narwhal class to read it if it uses a limited set of narratives. Using Narwhal to read hate speech would seem to be a natural application. Hope so, cuz it looks like the internet needs it.

I am up late reading about "AI and Hate Speech" (my Google search terms) and about something that removes 90% of the offending speech. That is probably as good as it will ever get, being model-less.

(A few minutes later after reading more) It looks way more complicated than simply finding keywords and narrative patterns. So perhaps my "hope" is absurd.