Emerging Rumor Identification for Social Media with Hot Topic Detection

Abstract

A rumor is commonly defined as a statement whose true value is unverifiable. As rumor can spread misinformation around people, causing social problems such as panic, and the rapid growth of online social media has made it possible for rumors to spread more quickly, it is important to automatically identify rumors for social media. Existing methods on rumor detection always concentrate on telling rumor from truth with handcrafted regular expressions, dealing with out of date rumor related message. To solve this problem, we introduce a novel hot topic detection method combining bursty term identification and multi-dimension sentence modeling to automatically detect emerging hot topics for rumor identification. We conduct a comprehensive set of experiments on two data sets from real-world social media. Experiment results show that our emerging rumor identification for social media with hot topic detection work well both in news data set and twitter data set, and combining the hot topic detection with the rumor detection is possible to finish real-time rumor identification. We believe our method to automatically detect rumor will open new dimensions in analyzing online misinformation and other aspects of social media mining.