I'd almost be tempted to do a dumb deep learning thing just to guess how likely the contents of a textbox are to be a pronoun or not just to make this something moderatable, but ah, coursework.
Some programming subreddit recently added pronoun flairs and someone has this/self pronouns, which would probably get flagged as taking the piss but is pretty nifty if you ask me
Well the point isn't to make the AI decide whether you're taking the piss, it just decides whether there's a reasonable likelihood you're totally taking the piss (i.e. just writing fuck/off as your pronouns) and holding off on assignment until a moderator manually confirms. Upon doing so, the pronouns are added to a whitelist and it never asks again for that specific set of pronouns.
I'd almost be tempted to do a dumb deep learning thing just to guess how likely the contents of a textbox are to be a pronoun or not just to make this something moderatable, but ah, coursework.
Some programming subreddit recently added pronoun flairs and someone has
this/selfpronouns, which would probably get flagged as taking the piss but is pretty nifty if you ask meWell the point isn't to make the AI decide whether you're taking the piss, it just decides whether there's a reasonable likelihood you're totally taking the piss (i.e. just writing fuck/off as your pronouns) and holding off on assignment until a moderator manually confirms. Upon doing so, the pronouns are added to a whitelist and it never asks again for that specific set of pronouns.
That would actually be a neat function for an automod to do. You could basically run spam filter heuristics so it can continuously get better