Rumor Detection in Twitter with Social Graph Structures

Since many people are often confused by rumors diffused in Twitter, it is required to detect them effectively. In this paper, we attempt to make good use of social graph structures in the detection. We study features derived from the structure which has been studied in other application fields and consider if they could be useful for rumor detection. We empirically investigate if they are effective for the detection under a support vector machine classifier. Experimental results show that the structures could be useful for improving detection performance. Our observations with J48 decision tree indicate that users' purposes in using Twitter inferred from social graph structures could relate to information credibility of tweets.