Business Intelligence According to Aspect-Based Sentiment Analysis using Double Propagation

Abstract

Market research is necessary to plan a game. Given game reviews, response from gamers in terms of aspect and sentiment expressions, can be extracted automatically by aspect-based sentiment analysis (ABSA). This paper develops business intelligence according to the result of double propagation, an ABSA method, to provide insight for game planning. Double propagation extracts pairs of aspect and sentiment expressions, and aggregates them into aspect category. For our corpus, DP gained F-Measure of 0.6239 for extracting aspect expression, F-Measure of 0.6239 for extracting sentiment expression, and F-Measure of 0.5136 for aspect aggregation. We also applied ETL (Extract Transform Load) to transform extraction result into data mart, and employ Tableau for data visualization. The aspect expression results displayed on data visualization giving the ratio information of positive and negative sentiment, which types of aspect categories game player interested in, and which types of targets needed to be concerned in designing game.