Representative example of words belonging to clusters obtained by TENMF before and after Japanese earthquake.

To conduct the experiment, we collect tweets from 4th to 16th March 2011: the period for the huge earthquakes in Japan. Clusters represent some characteristics of user groups. We list the words with high scores that might represent the characteristics of each cluster. Italic words here are originally in Japanese shown in parentheses, and translated into English by the authors. If further parentheses are attached, the words are explained. Cluster 1 can be interpreted as ‘English speaking’ users or ‘symbols’ everyone uses. Cluster 2 can represent ‘businesspeople’. Cluster 3 can represent ‘Internet addicted’ users. Cluster 4 can represent ‘news sources’.Representative example of words belonging to clusters obtained by TENMF before and after Japanese earthquake.