This project predicts whether or not a Twitter user is a troll using NLP on recent tweets.
The website for this project is http://www.trolltrackr.com. There you can go, enter the screen name for a Twitter user and get back a prediction about trolling behavior, and any recent trolling activity.
To create the model, I gathered tweets using the work in CollectingTweets.ipynb and then created the model in Finding_Trolls_and_Building_Models.ipynb.
To use this work in your own project, clone this repo and add a config.py with your Twitter credentials (go to https://apps.twitter.com/ to get some!). Then you put those credentials in a file config.py.