Discovering Social Bursts by Using Link Analytics on Large-Scale Social Networks

Abstract

Social Network Services (SNSs) have been regarded as an important source for identifying events in our society. Detecting and understanding social events from SNS has been investigated in many different contexts. Most of the studies have focused on detecting bursts based on textual context. In this paper, we propose a novel framework on collecting and analyzing social media data for i) discovering social bursts and ii) ranking these social bursts. Firstly, we detect social bursts from the photos textual annotations as well as visual features (e.g., timestamp and location); and then effectively identify social bursts by considering the spreading effect of social bursts in the spatio-temporal contexts. Secondly, we use these relationships among social bursts (e.g., spatial contexts, temporal contexts and content) for enhancing the precision of the algorithm. Finally, we rank social bursts by analyzing relationships between them (e.g., locations, timestamps, tags) at different period of time. The experiments have been conducted with two different approaches: i) offline approach with the collected dataset, and i i ) online approach with the streaming dataset in real time.