SMAM: Detecting Rumors from Microblogs with Stance Mining Assisting Task

With the rapid development of the Internet, the Microblogging platform has become an important social media as well as an ideal platform for rumor spreading. Existing approaches mainly rely on the hand-crafted features which require daunting manual effort. Deep learning methods require less manual effort but none of them take full advantage of stance information. Stance information is an important feature for rumor detection because users exposed to rumors tend to express more querying and denying stances. A Stance Mining Assisting Model (SMAS) making full use of stance information is proposed in this paper. First, all the posts in a microblog event are divided into N subevents. Second, two deep learning network structures are proposed to learn hidden representations about stance mining and rumor detection respectively. Third, two hidden representations are concatenated into an integrated representation to detect rumor. The experimental results on dataset from Weibo platforms demonstrate that SMAS is effective in rumor detecting and the accuracy of our model achieves 93.9%.