Call Attention to Rumors: Deep Attention Based Recurrent Neural Networks for Early Rumor Detection

Œe proliferation of social media in communication and information dissemination has made it an ideal platform for spreading rumors. Automatically debunking rumors at their stage of di‚usion is known as early rumor detection, which refers to dealing with sequential posts regarding disputed factual claims with certain variations and highly textual duplication over time. Œus, identifying trending rumors demands an ecient yet ƒexible model that is able to capture long-range dependencies among postings and produce distinct representations for the accurate early detection. However, it is a challenging task to apply conventional classi€cation algorithms to rumor detection in earliness since they rely on hand-cra‰ed features which require intensive manual e‚orts in the case of large amount of posts. Œis paper presents a deep aŠention model on the basis of recurrent neural networks (RNN) to learn selectively temporal hidden representations of sequential posts for identifying rumors. Œe proposed model delves so‰-aŠention into the recurrence to simultaneously pool out distinct features with particular focus and produce hidden representations that capture contextual variations of relevant posts over time. Extensive experiments on real datasets collected from social media websites demonstrate that (1) the deep aŠention based RNN model outperforms state-of-thearts that rely on hand-cra‰ed features; (2) the introduction of so‰ aŠention mechanism can e‚ectively distill relevant parts to rumors from original posts in advance; (3) the proposed method detects rumors more quickly and accurately than competitors.