Online indexing and clustering of social media data for emergency management
Social media becomes a vital part in our daily communication practice, creating a huge amount of data and covering different real-world situations. Currently, there is a tendency in making use of social media during emergency management and response. Most of this effort is performed by a huge number of volunteers browsing through social media data and preparing maps that can be used by professional first responders. Automatic analysis approaches are needed to directly support the response teams in monitoring and also understanding the evolution of facts in social media during an emergency situation. In this paper, we investigate the problem of real-time sub-events identification in social media data (i.e., Twitter, Flickr and YouTube) during emergencies. A processing framework is presented serving to generate situational reports/summaries from social media data. This framework relies in particular on online indexing and online clustering of media data streams. Online indexing aims at tracking the relevant vocabulary to capture the evolution of sub-events over time. Online clustering, on the other hand, is used to detect and update the set of sub-events using the indices built during online indexing. To evaluate the framework, social media data related to Hurricane Sandy 2012 was collected and used in a series of experiments. In particular some online indexing methods have been tested against a proposed method to show their suitability. Moreover, the quality of online clustering has been studied using standard clustering indices. Overall the framework provides a great opportunity for supporting emergency responders as demonstrated in real-world emergency exercises.
65 references, page 1 of 7
views in local repository
downloads in local repository
The information is available from the following content providers: