A friend posted a humorous photo on Facebook the other day of a sign that read “In Case of Fire… Exit building before Tweeting about it.” Hopefully it’s a made-up sign just to get a laugh, but the truth is actually closer than you think. The intersection of the pervasiveness of mobile devices and the ubiquitous nature of social media is creating a tremendous amount of relevant data about the physical world not previously seen before – what is happening, where it’s happening, to whom, by whom in some cases and when. Not usually early adopters, public safety officials are starting to take notice.
Here’s a real-world story:
In 2009 in Adelaide, Australia, two young girls trapped in a storm drain that was filling quickly with rain water updated their Facebook status instead of calling emergency services (Mashable: Trapped Girls Updated Facebook Status Instead of Calling for Help). Very lucky for them, a friend was online, saw the status update and called emergency services with the information saving the girls before the situation could have turned worse.
Organizations such as the U.S. Centers for Disease Control (CDC), Singapore’s Ministry of Health, and two Brazilian National Institutes of Science and Technology have started investigating how they can continuously monitor social media sites for early warning detection of health outbreak issues. This can include tweets about SARS infections, postings of dengue fever symptoms occurring in family members, or even sharing a photo of a rabid dog in a playground. First responders such as fire departments and emergency medical personnel are keenly interested in early warning detection, too; any minute sooner that they respond can literally mean the difference in saving a life.
From a technology perspective, keyword monitoring and sentiment analysis of social media is a first (and easiest) step in identify early issues. The harder part is how you can analyze those mentions for a temporal and geospatial component which not only localizes social postings to a geographic area (city, street address, latitude/longitude coordinates), but also help detect the change in rate at which social mentions are occurring. A single mention weeks apart is not as interesting (or relevant) as five mentions from five different people within a ten minute window in the same geographic area. This rate change is the essence of early warning detection.
Complex Event Processing (CEP) is software technology that discovers just that: events of interest (posts, tweets) occurring with certain attributes (keywords, sentiment) within a certain time window (temporal, rate) near a specific location (geospatial). CEP is the secret technology sauce that is little-known to users outside of its core markets of intelligence, military, supply chain and financial services. In other markets, it has the opportunity to change how organizations detect and respond with relevant and timely actions that can give them a competitive advantage.