Automatic detection and verification of rumors on Twitter

The spread of malicious or accidental misinformation in social media, especially in timesensitive situations such as real-world emergencies can have harmful effects on individuals and society. This thesis develops models for automated detection and verification of rumors (unverified information) that propagate through Twitter. Detection of rumors about an event is achieved through classifying and clustering assertions made about that event. Assertions are classified through a speech-act classifier for Twitter developed for this thesis. The classifier utilizes a combination of semantic and syntactic features to identify assertions with 91% accuracy. To predict the veracity of rumors, we identify salient features of rumors by examining three aspects of information spread: linguistic style used to express rumors, characteristics of people involved in propagating information, and network propagation dynamics. The predicted veracity of a time series of these features extracted from a rumor (a collection of tweets) is generated using Hidden Markov Models. The verification algorithm was tested on 209 rumors representing 938,806 tweets collected from real-world events including the 2013 Boston Marathon bombings, the 2014 Ferguson unrest and the 2014 Ebola epidemic, and many other rumors reported on popular websites that document public rumors. The algorithm is able to predict the veracity of rumors with an accuracy of 75%. The ability to track rumors and predict their outcomes may have practical applications for news consumers, financial markets, journalists, and emergency services, and more generally to help minimize the impact of false information on Twitter. Thesis Supervisor: Deb K. Roy Title: Associate Professor, Program in Media Arts and Sciences