Time-series Keyword Extraction Method and Its Application to Discovery Japanese Key Technology Transition Insights

Abstract

This paper presents a time-series keyword extraction method and its application to the discovery of key technological transition insights in Japan. In general, it is one of the most important issues in understanding trends in science and technology. To grasp these trends, it will be possible to visualize trends in science and technology from time to time if a method is established to extract important keywords from time to time using the White Paper on Science, Technology, and Innovation published every year by the Japanese government as an example. In this method, words are extracted from text data from time to time, and a new Importance Transition Discovering Score (WITD-Score) is proposed as an index representing the likelihood of the occurrence of each word at that time, following the concept of F-Score. By extracting the transition of keywords from time to time from the change in the WITD-Score for each word, we can extract the transition of keywords from time to time. By implementing this method, we can discover important keywords from time-series text data and visualize the transition of keywords.