PPTweets - Spawned by Demon Hacks

Try it out! - pptweets.com

Social media manipulation, trolling, and agenda pushing have become hot button issues over the past year, affecting American society all the way from teen bullying up to the presidential election. An unchecked and unmanaged environment can become poisonous when civil conversation breaks down, an occurrence which is increasing in frequency now that there are people and organizations putting in a concerted effort to make sure this is the case.

A major issue with managing this problem is identifying the source of such vitriol and malcontent to begin with - social media networks are by definitions distributed, and it’s tough to track down the trouble makers when report buttons are abused by users or ignored by the networks themselves.

Poison Pill Tweets is a proof of concept for a kind of “stack trace for hatred”, a tool that tracks controversial tweets back to the source. Using natural language processing to score tweets, we can quantify the level and magnitude of negativity (i.e. how negative the content, and how emphatic the poster) and determine what has kicked off the negative chain.

By analyzing the user profile responsible for an incendiary tweet, we can determine if they are regularly the source of such negativity and someone who may be flagged, warned, and possibly banned.


Share this project: