Potential Information Maximization: Potentiality-Driven Information Maximization and Its Application to Tweets Classification and Interpretation

Abstract

The present paper aims to apply a new informationtheoretic learning method called “potential information maximization” to the classification and interpretation of tweets. It is well known that social media sites such as Twitter play a crucial role in transmitting important information during natural disasters. In particular, since the Great East Japan Earthquake in 2011, Twitter has been considered as one of the most efficient and convenient communication tools. However, since there is much redundant information contained in tweets, it is critical that methods be developed to extract only the most important information from them. To cope with complex and redundant data, a new neural information-theoretic learning method has been developed for this purpose. The method aims to find neurons with high potential and maximize their information content to reduce redundancy and to focus on important information. The method was applied to real tweet data collected during the earthquake. It was found that the method could classify the tweets as important and unimportant more accurately than other conventional machine learning methods. In addition, the method made it possible to interpret how the tweets could be classified based on the examination of highly potential neurons.