Monitoring Spatial Coverage of Trending Topics in Twitter

Most messages posted in Twitter usually discuss an ongoing event, triggering a series of tweets that together may constitute a trending topic (e.g., #election2012, #jesuischarlie, #oscars2016). Sometimes, such a topic may be trending only locally, assuming that related posts have a geographical reference, either directly geotagging them with exact coordinates or indirectly by mentioning a well-known landmark (e.g., #bataclan). In this paper, we study how trending topics evolve both in space and time, by monitoring the Twitter stream and detecting online the varying spatial coverage of related geotagged posts across time. Observing the evolving spatial coverage of such posts may reveal the intensity of a phenomenon and its impact on local communities, and can further assist in improving user awareness on facts and situations with strong local footprint. We propose a technique that can maintain trending topics and readily recognize their locality by subdividing the area of interest into elementary cells. Thus, instead of costly spatial clustering of incoming messages by topic, we can approximately, but almost instantly, identify such areas of coverage as groups of contiguous cells, as well as their mutability with time. We conducted a comprehensive empirical study to evaluate the performance of the proposed methodology, as well as the quality of detected areas of coverage. Results confirm that our technique can efficiently cope with scalable volumes of messages, offering incremental response in real-time regarding coverage updates for trending topics.