Volunteer-powered automatic classification of social media messages for public health in AIDR

Microblogging platforms such as Twitter have become a valuable resource for disease surveillance and monitoring. Automatic classification can be used to detect disease-related messages and to sort them into meaningful categories. In this paper, we show how the AIDR (Artificial Intelligence for Disaster Response) platform can be used to harvest and perform analysis of tweets in real-time using supervised machine learning techniques. AIDR is a volunteer-powered online social media content classification platform that automatically learns from a set of human-annotated examples to classify tweets into user-defined categories. In addition, it automatically increases classification accuracy as new examples become available. AIDR can be operated through a web interface without the need to deal with the complexity of the machine learning methods used.