An insight analysis and detection of drug-abuse risk behavior on Twitter with self-taught deep learning

Computational Social Networks

Table 1 Instances of manually annotated positive tweets and negative tweets

	Tweets
Positive	“Ever since my Acid trips like whenever I get super high I just start—lightly hallucinating and it’s tbh creepy”
	“drove like 10 miles on these icy ass roads all to get some weed if imma—be locked up in my house for awhile imma need some weed”
	“Smoking a blunt at home so much better than going to the woods in—Brooksville and puking on yourself Bc you drank too much reball”
Negative	“Just watched Fear and Loathing in Las Vegas for the first time—and I think I should have been on acid to fully understand it”
	“today I was asked if I do heroin because I went to Lancaster????”
	“Morgan told me my Bitmoji looks like a heroin addict?”