Tadashi Yagi, Seiya Murata

(Doshisha University, Japan)

 

 

Abstract: We present a method for analyzing the determinants of ratings in reviewing creative works by examining the case of movies in Japan. In Japan, the Yahoo Movie Review website presents viewer ratings, comments, the number of followers, and evaluations by followers. We develop a system for extracting information from this website and analyzing keywords in comments for positive and negative emotional expressions by using the “IBM SPSS Text Analytics for Survey” text-mining software. This allows conversion of text data to numerical data, and application of principal factor analysis and regression analysis. We generate variables representing evaluations of “plot”, “cultural value”, “main theme”, and “setting” of each movie, and analyze the sign, positive or negative, of each variable’s coefficient in a rating by regression analysis. We present a case showing positive evaluations for plot, cultural value, and setting, but negative evaluation for theme.

 

 

Key words: text mining; movie review; big data; evaluation

 

JEL code: Z1