Automatic detection of rumor on Sina Weibo

Abstract

The problem of gauging information credibility on social networks has received considerable attention in recent years. Most previous work has chosen Twitter, the world's largest micro-blogging platform, as the premise of research. In this work, we shift the premise and study the problem of information credibility on Sina Weibo, China's leading microblogging service provider. With eight times more users than Twitter, Sina Weibo is more of a Facebook-Twitter hybrid than a pure Twitter clone, and exhibits several important characteristics that distinguish it from Twitter. We collect an extensive set of microblogs which have been confirmed to be false rumors based on information from the official rumor-busting service provided by Sina Weibo. Unlike previous studies on Twitter where the labeling of rumors is done manually by the participants of the experiments, the official nature of this service ensures the high quality of the dataset. We then examine an extensive set of features that can be extracted from the microblogs, and train a classifier to automatically detect the rumors from a mixed set of true information and false information. The experiments show that some of the new features we propose are indeed effective in the classification, and even the features considered in previous studies have different implications with Sina Weibo than with Twitter. To the best of our knowledge, this is the first study on rumor analysis and detection on Sina Weibo.

Detect Rumors in Microblog Posts Using Propagation Structure via Kernel Learning

“…DTC and SVM-RBF: The Twitter information credibility model using Decision Tree Classifier (Castillo et al, 2011) and the SVM-based model with RBF kernel (Yang et al, 2012), respectively, both using hand-crafted features based on the overall statistics of the posts.…”

“…Castillo et al (2011) studied information credibility on Twitter using a wide range of hand-crafted features. Following that, various features corresponding to message contents, user profiles and statistics of propagation patterns were proposed in many studies (Yang et al, 2012;Wu et al, 2015;Sun et al, 2013;Liu et al, 2015). Zhao et al (2015) focused on early rumor detection by using regular expressions for finding questing and denying tweets as the key for debunking rumor.…”

“…However, such endeavor is manual, thus prone to poor coverage and low speed. Feature-based methods (Castillo et al, 2011;Yang et al, 2012;Ma et al, 2015) achieved certain success by employing large feature sets crafted from message contents, user profiles and holistic statistics of diffusion patterns (e.g., number of retweets, propagation time, etc.). But such an approach was over simplified as they ignored the dynamics of rumor propagation.…”

Deception Detection and Rumor Debunking for Social Media

“…Most existing algorithms for debunking rumors, however, follow Castillo, Mendoza, and Poblete's work (Castillo et al, 2011, Mendoza et al, 2010 employing variations on data used and features extracted (Wu et al, 2015, Yang et al, 2012. Qazvinian and colleagues (2011) focus on rumor-related tweets to match certain regular expression of the keyword query and the users' believing behavior about those rumor-related tweets; both pieces of information are instrumental in isolating rumors.…”

“…Qazvinian and colleagues (2011) focus on rumor-related tweets to match certain regular expression of the keyword query and the users' believing behavior about those rumor-related tweets; both pieces of information are instrumental in isolating rumors. Mendoza and colleagues (2010) analyze user behavior through tweets during the Chilean earthquake that year: "they analyze users' retweeting topology network and the difference in the rumor diffusion pattern on Twitter environment than on traditional news platforms" (Yang et al, 2012). Moving away from Twitter, Yang and colleagues (2012) studied Sina Weibo, China's leading micro-blogging service provider that functions like a FacebookTwitter hybrid.…”