[xXOpkillerXx]

Originally Posted by Arch0wl

had no idea this existed, no. completely unfamiliar with NLP as a field.

I'm concerned with the application to language rules / speech codes, such as the vagueness in policies that prohibit insults -- because for every policy that prohibits insults, it also seems to prohibt neutral, but displeasing information about other people.

Well, that would be a very hard task and here's why:

1. You have to define a logical structure from a grammar and most likely use some probalistic analysis on words. See this for info on NLP grammars. See that for probabilistic parsing.

2. When backing up your claim that x structure tends to be content that prohibits insults, you Must stay objective in your definition of "prohibiting insults". You also need to somehow label many policies (prohibits insults y/n, prohibits neutral displeasing y/n, etc) or find such a dataset. Labeling that is a massive task because you can probably hardly remain objective, unless you come up with a thorough labeling system that is accepted.

3. Same thing kinda for vagueness; that is probably a badly defined metric of language and would need a strong logical definition (in terms of NLP grammar, at least).