Real-time Event Detection on Social Data Stream

Social Network Services (SNS) are becoming more popular in our daily life, the process is boosted by various kinds of smart devices integrating utility modules such as 3G/WIFI connector, GPS tracker, Camera, Heartbeat sensor and so on. It makes the information flow (or Social Data Stream) on SNS have a real-time nature characteristic, where each SNS user is an information sensor and also a data connector for diffusing interesting news to his/her communication networks. Hiding inside the information flow are pieces of real social events. The events draw attention from users evidencing by the number of relevant announces and communication interactions toward that topic. However, traditional topic detection approaches are not designed to detect the kind of the event efficiently in real-time, particularly if the data sources are influenced by noise data and containing diverse topics. To overcome the issue, in this paper we proposed a model for extracting and tracking real social events on Social Data Stream, which can work well in real-time by using distributing computation and data aggregation technique on the discrete signals as a new representation of the original data.