Closed corcra closed 2 years ago
For part #1: I'd like a list of 20 known trolls and 20 known non-trolls. Obviously this is a tiny data set for training purposes, but right now I'm not sure how to automatically acquire high-quality data like this. Ideas welcome.
We need to train our algorithm to identify trolls. (
calculate_troll_score
intweep.py
)This can either be hand-devised with rules, or we can try to learn it using a database of known trolls. Maybe we want both.