That seems really hard. How would you identify a joke/insult vs a neutral post vs praise?
Lots of correlation models that need to be updated as people learn what gets logged as certain sentiment and then they change their language to not be part of that.
Then I realized you might be asking about "how do you identify a sarcastic comment that looks like praise?" ..and it seems there's no easy answer to that.
https://journals.sagepub.com/doi/10.1177/2053951720972735
https://wpmu.mah.se/nmict162group7/?p=180
So I guess after all this, I've gone from my initial sarcastic remark (the top of this post) to 10 mins later scratching my head and thinking that this business idea is more difficult that going back to our government and demanding more change.