A method to quantify twitter user´s posting activities for constructing disaster information support system

In Twitter, there have been various kinds of Tweets and also the total volume is extremely huge, so some kind of effectively filterable technique in response to the contents of Tweets or each user´s purpose is considered to be essential in order for Twitter to be effective at the time of disaster. A framework for study aiming to effectively utilize social media at the time of disaster was proposed in which some kind of filtering mechanism to easily catch real human´s voices was considered to be important for getting better performance out of social media. This paper focuses on a method to numerically-represent the characteristics of Twitter users in response to each user´s tweeting, replying, and retweeting activities, which are assumed to be the source of Twitter´s real time feature, and the characteristics depending on Twitter user´s attributes were clarified on the basis of the multidimensional values of entropies by using the real Tweets distributed at the time of the Great East Japan Earthquake. Also by modifying this method, this paper newly proposes a dynamic quantification method. The method is so designed as to be able to dynamically change the analysis period. From the experiment, the proposed method can filter bot users comparable to the existing contents based filtering method.