make an updateable slang DB, tie it to knowyourmeme and other sources, have it extract to a vector db for use when prompting the model.
now it stays up-to-date and you correct bad translations. it would be capable of translation as well as using the encoding sets in any way you can think of.
Probably don’t even need AI for that. English slang is extremely simple in that it mostly replaces or abbreviates discrete words or phrases, so you can pretty much just find and replace and be able to decode most of it. Not perfect, but if you’re telling a human what it means they can probably figure out any remaining ones that were a bit more complex.
make an updateable slang DB, tie it to knowyourmeme and other sources, have it extract to a vector db for use when prompting the model.
now it stays up-to-date and you correct bad translations. it would be capable of translation as well as using the encoding sets in any way you can think of.
Probably don’t even need AI for that. English slang is extremely simple in that it mostly replaces or abbreviates discrete words or phrases, so you can pretty much just find and replace and be able to decode most of it. Not perfect, but if you’re telling a human what it means they can probably figure out any remaining ones that were a bit more complex.