INTEGRATING TOPIC MODEL AND DISCRETE-CONTINUOUS MODEL TO ANALYZE FREE ANSWER DATA

Abstract

Text mining approach for analyzing free answer data attracts attention in infrastructure and city planning studies. This paper proposed a novel framework integrating topic model and discrete-continuous model to analyze free answer data. We use the interview survey data for households in Mashiki temporary housings following the 2016 Kumamoto earthquake. Topic model describes the ratio of topic in interview for each respondent and discrete-continuous model describes the topic choice and the ratio of topic. Our new framework statistically demonstrates the relationship between respondent attributes and topic in interview.