Extracting Transportation Information and Traffic Problems from Tweets during a Disaster

In a disaster, one of the most important issues for victims is how to find evacuation routes to safety from hazardous areas. To offer such routes, we propose methods automatically extracting transportation information and traffic problems from tweets written in Japanese and posted during a disaster. To investigate the effectiveness of our methods, we conducted some experiments using tweets posted during the Great Eastern Japan Earthquake in March 2011. From the experimental results, we obtained precision of 78.2% and recall of 53.4% in automatic extraction of transportation information. For extracting traffic problems, we identified tweets containing relevant information (we call them traffic problem tweets), and extracted traffic problem from them. In identifying traffic problem tweets, we obtained precision of 77.7% and recall of 70.7%. In extracting traffic problems, we obtained precision of 87.0% and recall of 57.1%. Thus, we have constructed a system for providing transportation information and traffic problems in a disaster. Keywords-disaster; evacuation routes; information extraction;