Exploring Scalability and Time-Sensitiveness in Reliable Social Sensing with Accuracy Assessment

This paper presents a scalable estimation theoretic framework to address the time-sensitive truth discovery problem with accuracy assessment in social sensing applications. Social sensing has emerged as a new application paradigm that provides us with an unprecedented opportunity to collect observations about the physical world from humans or devices on their behalf. A fundamental challenge in social sensing applications lies in ascertaining the correctness of claims and the reliability of data sources without knowing either of them a priori, which is referred to as truth discovery. While significant progress has been made to solve the truth discovery problem, there exists three important limitations: 1) The information of users and claims in time dimension has not been fully exploited in the truth discovery solutions; 2) An analytical framework to rigorously assess the accuracy of the truth discovery results is lacking; and 3) Many current truth discovery schemes perform sequential operations, which are not scalable to large-scale social sensing events. To address the above limitations, we propose a scalable time-sensitive truth discovery (TS-TD) scheme that explicitly incorporates the source responsiveness and the claim lifespan into an estimation theoretical framework. Furthermore, we develop new confidence bounds to rigorously assess the accuracy of the truth discovery results. We also implement a parallel TS-TD algorithm on a graphic processing unit platform with thousands of cores to improve the computational efficiency. Finally, we evaluate the TS-TD scheme through three real-world case studies using Twitter data feeds and a simulation study. The evaluation results demonstrate the effectiveness and efficiency of our scheme.