Extracting and Exploring Information about Flood Events from Twitter

Abstract: The more information about a disaster gets established, the more efficiently the disastermanagement is done by the concerned parties to handle the situation. People tend to sharetheir experiences during disastrous events using social media, making them potential datasources. This thesis project implements a pipeline to extract knowledge from Twitter about floodevents. It determines flood-relevant tweets using a classifier and identifies geographicallocations mentioned in the tweets using a hybrid geoparsing approach. At the end of thepipeline, the spatial, temporal, and textual aspects of the results are presented using aninteractive visual interface. The implemented pipeline is exemplified using historical tweetscreated during past flood events