Author identification of Japanese texts based on notations and syllabic writings usage patterns

Abstract

　This paper presents a novel feature to discriminate authors from Japanese texts, based on notations and syllabic writings usage patterns. These two properties are particular to Japanese texts. The proposed indicator's performance was compared to part-of-speech 3-gram and comma usage patterns, both considered as best methods for author identification tasks. Texts written by college students and published novels were analyzed. The proposed indicator provides a novel viewpoint to capture authors' characteristics, and its precision is equivalent to conventional best indicators.