SportSense: Real-Time Detection of NFL Game Events from Twitter

Abstract: We report our experience in building a working system, SportSense (http://www.sportsense.us), which exploits Twitter users as human sensors of the physical world to detect events in real-time. Using the US National Football League (NFL) games as a case study, we report in-depth measurement studies of the delay and post rate of tweets, and their dependence on other properties. We subsequently develop a novel event detection method based on these findings, and demonstrate that it can effectively and accurately extract game events using open access Twitter data. SportSense has been evolving during the 2010-11 and 2011-12 NFL seasons and is able to recognize NFL game big plays in 30 to 90 seconds with 98% true positive, and 9% false positive rates. Using a smart electronic TV program guide, we show that SportSense can utilize human sensors to empower novel services.