Abstract
After the Great East Japan Earthquake, which occurred on March 11, 2011, most of the communication systems stopped because of power outages, and social networking services (SNS) such as Twitter made significant contributions as communication tools. A lot of information was transmitted on Twitter and details such as who needed help where, and what was lacking could be known. The government could also determine what was happening in the affected area in real time from tweet data. On the other hand, many false remarks and rumors were also circulated on Twitter. As a result, all information had to be checked manually, and it took long time before rescue parties and aid supplies could be sent. In this study, we propose how to find key words related to disasters automatically using a naïve Bayesian classifier and information value. Using this method, we show that it is possible to determine the current status of the affected area in real time. Additionally, we compare the naïve Bayesian classifier with a support vector machine (SVM) and show that the naïve Bayesian classifier can classify tweet data to the same accuracy level as that of the SVM.
