Adaptive information extraction of disaster information from Twitter

Abstract

With the popularity of the Internet and social media platforms, information that is potentially useful in disaster response becomes available online in the hours and days immediately following a disaster. The use of information extraction in retrieving relevant disaster information from all these crowdsourced data would provide more information coming from both official reports, and the affected people themselves which in turn facilitate better decision making environments for disaster managers. This paper describes a system which performs an adaptive information retrieval of disaster related information coming from Twitter. Result shows 94.33% accuracy when extracting disaster and location information in the typhoon corpus while 90.79% accuracy for the fire corpus.