Arabic Rumor Detection using Collaborative Framework over Social Networks and Web

Abstract

The popularity of social networks and the fast propagation of information on such networks have increased their importance to be the main source of information for people. Any user has an account in social networks can be easily a reporter through posting information. However, social networks dont provide a service to check the truth of the posted information, which makes social networks users not distinguish between truth and untruth. Indeed, the absence of verifying information service may cause a propagation of fake information, so-called ”Rumor”. Which may lead to many serious problems. The researchers addressed the rumor problem through examining the propagation behavior of rumor information by giving the most attention to detect rumor topics instead of verifying individual posts. However, the use of graph metrics in studying the propagation behavior is unsuitable for real time detection because of the high time complexity. Also, examining the propagation behavior of information prevents the early detection of rumors. In this paper, we intend to design a system that detects false information. Our approach differs from others in treating the problem from the information retrieval point of view. More precisely, we perform collaboration between a set of social networks and web resources of information through querying them using the desired post to be verified as query. Then, a set of operations are applied on the information retrieved from social networks and web resources, including date filtering, content semantic matching, user and content features extraction. Our collaborative framework is generalizable to verify the truth of either individual post or topic. Also, we design this framework to detect Arabic rumors as first work in this area, to the best of our knowledge. The proposed algorithms are effective in rumor detection on social network.