|Speaker :||Oana Balalau|
|Time:||4:00 pm - 4:30 pm|
|Location:||LINCS / EIT Digital|
Social media contain information shared by hundreds of millions of users across the world and provide one of the richest dataset created by human activity. Social media distinguish themselves from traditional media as users can follow news channels, but they can also be the ones reporting updates on events.
In our research, we focus on the detection of disaster events, such as earthquakes, terrorist attacks, mass shootings, air strikes, etc. We develop tools to extract a useful description of an event via a graph-based approach and we show that our algorithm can incorporate information from several sources, such as both social and mainstream media.
Complementary to our work on event detection, we study several definitions of meaningful subgraphs in a graph. Density is a very good measure of importance and cohesiveness in graphs and dense subgraphs have been found to represent communities in social networks, or to indicate the presence of real-world events in a graph-of-words. We investigate both semantic and structural properties of dense subgraphs and we develop algorithms that can successfully scale to real-world graphs.