Gleaning valuable data from social platforms such as Twitter—particularly to map out critical location information during emergencies— has become more effective and efficient thanks to Oak Ridge National Laboratory. In a preliminary study, geospatial scientists at ORNL built a classification model that can quickly collect, sift through and store large amounts of data, including relevant text, sensor data and images, and automatically detect where a power outage has occurred. “For this data to be useful in an emergency, it needs to be validated and made available to utility companies or first responders in real time,” said ORNL’s Gautam Thakur. “We developed algorithms through deep learning methods, ran them on customized hardware and mechanisms designed by ORNL and successfully extracted outage and location information from informal, social media text in near-real-time.” This demonstration focused on power outages, but the method could be extended to other disaster analysis.
Topic: National Security