Human as Real-Time Sensors of Social and Physical Events: A Case Study of Twitter and Sports Games

In this work, we study how Twitter can be used as a sensor to detect frequent and diverse social and physical events in real-time. We devise efficient data collection and event recognition solutions that work despite various limits on free access to Twitter data. We describe a web service implementation of our solution and report our experience with the 2010-2011 US National Football League (NFL) games. The service was able to recognize NFL game events within 40 seconds and with accuracy up to 90%. This capability will be very useful for not only real-time electronic program guide for live broadcast programs but also refined auction of advertisement slots. More importantly, it demonstrates for the first time the feasibility of using Twitter for real-time social and physical event detection for ubiquitous computing.