The Social Media Tracking and Analysis System (SMTAS) has contributed to the acquisition of a recent grant funded by the National Oceanic and Atmospheric Administration through the Coastal Storms Awareness Program, managed by the New York, New Jersey and Connecticut Sea Grant Consortium. A team of social scientists and computer programmers at Mississippi State University received funding for this research through the Connecticut Sea Grant program. The title of the project is: Assessment of Social Media Usage during Severe Weather Events and the Development of a Twitter-based Model for Improved Communication of Storm-related Information. One of the main goals of the project is to identify key impact factors affecting the dissemination of storm-related information.
To that end, SMTAS was utilized to collect approximately 12 million tweets during Hurricane Sandy. In order to understand the networks of key-users and the topics that were discussed during the event, researchers utilized the metadata attributes of the raw Twitter data such as user mentions, re-tweet data, and tweet content to analyze/visualize the inter-connectivity of tweets and their users in a graph based model.
The first graph (shown below), is the visualization of connected (users who have either mentioned or re-tweeted another user) users during Hurricane Sandy. In the graph, each node is an individual Twitter user. The size of the node is representative of the frequency of mentions or re-tweets. An edge/line between two nodes exists if there is re-tweet or mention connectivity. The colors of the edges are based on the Modularity index of a node cluster (i.e. the connectivity of a cluster of nodes). The analysis is able to identify prominent users that were mentioned during the time of the event and their closely associated Twitter user. The presence of news agencies (cnnbrk, NewYorkPost, mashable, HuffingtonPost), politicians (MikeBloomberg, BarackObama, GovChristie, CoryBooker), fedral agencies (fema, NOAA), etc, and their connectivity to other users can be identified.
For dynamic graph please use the following link to download pdf: Link
The second graph shown below is the network of word co-existence in Twitter messages that were collected during Hurricane Sandy. Researcher are able to identify clusters of words which such as Sandy, Hurricane, Storm, which tend to co-exist together. Words such as Everyone, Stay, Safe, presents another unique topic. Power, still, out, suggests tweets that were talking about the power outages. The network visualization is also able to differentiate noisy tweets (words with you, never, love, LOL, etc) and tweets in other languages (spanish speaking community cluster in the bottom left).
For dynamic graph of word collation please use the following link to download pdf: Link