Modeling and Detecting Online Rumors
Abstract
Today, social media provide the means by which billions of people experience news and events happening around the world. Despite many organized attempts along this line of research, social media is still opaque to the credibility of news and information. When you view your social media feed, you have no sense as to which parts are reliable and which are not. This work addresses this gap, by building large-scale, generalizable science around credibility in social media. First, it offers an iterative framework for systematically tracking the credibility of social media information. Next, by running the framework for several months on the popular social networking site Twitter, we compiled a corpus (CREDBANK) with newsworthy topics, their associated tweets and corresponding credibility scores. We show that a parsimonious language model can predict the credibility of newsworthy topics with high accuracy. Finally, by investigating differences in temporal dynamics through the lens of collective attention, we demonstrate that recurring attentional bursts are correlated with lower credible events.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1510280
Entities
People
- Eric Gilbert
Organizations
- Army Contracting Command
- Georgia Tech Research Corporation
- United States Army