Feature Extraction of Comedy and Tragedy by Stylometry

Abstract

　”Happy-Ending”and ”Bad-Ending”are the most major categorizations in the narrative, but little work has been done on quantitative analysis about Happy-Ending works and Bad-Ending works. This study was made to extract Comedy and Tragedy in Otsuichi’s short stories. Firstly, the target datas were classified into three genres, and a new Japanese Sentiment Polarity Dictionary was created based on the thesaurus. Secondly, polar words in the text were extracted with Polarity Dictionary. In addition,we extract features of Happy-Ending works and Bad-Ending works by hierarchical cluster analysis and multiple regression analysis. As a result, this study showed that polar words categorized as ”happy”,”love”, and ”respect”in the last quarter of the story contributed to the Happy-Ending works. This study also showed that polar words categorized as ”sorrow”and ”sad”in the last quarter of the story contributed to the Bad-Ending works.