I find VADER very crude and fallible for sentiment, did they check its reliability on each of their messages?
e.g. it struggles on handling negation, hyperbole, ambiguity:
"I cannot believe how impossibly good my code is this week."
I've thought about doing something very similar to this in the past. Seeing everything in graphs is beautiful.
But seeing this, I get the feeling that there's not really much use for it. I'd already know most of the things the data uncover. But maybe that's just a matter of timescale? I can imagine me being surprised how things changed over many years (5+years)