| dc.description.abstract | As breaking news unfolds, social media has become the go-to platform to learn about the latest updates from journalists and eyewitnesses on the ground. The fact that anybody can post content in social media during these breaking news leads to posting and diffusion of unverified rumours, which in turn produces uncertainty and increases anxiety. Given the scale of social media contents, automation is key for effective detection of these rumours. In this paper we introduce a novel approach to rumour detection that learns from the sequential dynamics of reporting during breaking news in social media to detect rumours in new stories. Using five Twitter datasets collected during breaking news stories, we experiment with Conditional Random Fields as a sequential classifier that leverages context learnt during an event for rumour detection, which we compare with the state-of-the-art rumour detection system as well as other baselines. In contrast to existing work, our classifier does not need to observe tweets querying a piece of information to deem it a rumour, but instead we detect rumours from the tweet alone by exploiting context learnt during the event. Further, we experiment with homophily as a predictive feature for detecting rumours, i.e. setting forth the hypothesis that a user will be more likely to post a rumour if they follow users who posted or spread rumours in the past. Our classifier achieves state-of-the-art performance, beating competitive baselines as well as outperforming our best baseline with nearly 40% improvement in terms of F1 score. Our research proves the effectiveness of the consideration of the sequential nature of social media data and the usefulness of homophily as a feature for rumour detection. | en_US |
