Abstract
Mass media such as television broadcast were used to be major information sources for citizens at disasters. However, after the Great East Japan Earthquake in 2011, citizens have started using social media, such as Twitter and blog, for sharing information of the disaster-affected area. Consequently, a lot of information is exchanged on social and mass media at the time of the disasters. It is important to consider characteristics of each medium for effective disaster information transmission. The purpose of this study is to clarify the differences in topics and release timings of information among social and mass media at the time of the disaster. This study conducts a non-negative tensor factorization analysis of articles posted on Twitter, 2channel, blog, board, internet news and television broadcast to extract topics and analyze differences in topics among media and their time-series transition at the earthquake caused on 21th October 2016 in Tottori prefecture in Japan. As a result, the characteristics of media was detected and the process of spreading topics between media was clarified.
