Automatic Generation of Event Timelines from Social Data

Abstract

Over the past few years, social media has seen phenomenal growth and has become a very important source for getting real time updates from different parts of the world. While the notion of a trend usually reflects current events, the amount of information accumulated over a period of time can be used to provide another perspective for such events in the form of a timeline. In this paper, we present a technique that uses social information as relevance surrogates to generate an informative timeline. A core component is a variation of pseudo relevance feedback that is automatically generated using social data without external evidence. Finally, we describe the implementation of such technique and present evaluation results using a real-world data set.