Identifying local burstiness in a sequence of batched georeferenced documents

Abstract

One of the most interesting emerging topics in social media is the increase in the number of georeferenced documents. These documents include not only text data, but also posted time and location. People have been transmitting information regarding items and events they have witnessed in their daily lives and collecting information on objects of interest through georeferenced documents. Therefore, many researchers are directing their attention to extracting local topics and events from georeferenced documents. In this paper, we propose a novel location-based burst detection algorithm for identifying the burstiness of a keyword related to local topics and events in a sequence of batched georeferenced documents, composed of ordered georeferenced document sets. Burstiness is one of the simplest yet most robust criteria for extracting hot topics and events from a sequence of batched documents. Identifying the burstiness of a keyword related to local topics and events captures not only the peak periods of the trending topics and events, but also the localities at which they are occurring. To evaluate the proposed location-based burst detection algorithm, we used an actual sequence of batched georeferenced documents that were composed by crawling tweets posted on the Twitter site. The experimental results confirm that the proposed location-based burst detection algorithm can identify location-based bursts successfully.