Event Identification and Tracking in Social Media Streaming Data

Abstract

In recent years, the growing popularity and active use of social media services on the web have resulted in massive amounts of user-generated data. With these data available, there is also an increasing interest in analyzing it and to extract information from it. Since social media analysis is concerned with investigating current events around the world, there is a strong emphasis on identifying these evens as quickly as possible, ideally in real-time. In order to scale with the rapidly increasing volume of social media data, we propose to explore very simple event identification mechanisms, rather than applying the more complex approaches that have been proposed in the literature. In this paper, we present a first investigation along this motivation. We discuss a simple sliding window model, which uses shifts in the inverse document frequency (IDF) to capture trending terms as well as to track the evolution and the context around events. Further, we present an initial experimental evaluation of the results that we obtained by analyzing real-world data streams from Twitter.