Cross-genre comparison and pattern extraction for the features of utterances of story characters

Little quantitative research has been conducted on the detailed expressions of story characters. This study focuses on the utterances of story characters and constructs a fundamental dataset. To compare different story genres, five popular genres (adventure, battle, love, detective, and horror) in modern Japan entertainment culture were selected based on comic and game sales rankings. Approximately 26,000 utterances of story characters in the five genres were randomly selected. The selected utterances were categorized from the perspective of speech acts. Genre and character role comparisons based on a chi-square test and residual analysis clarified the tendencies of each genre. Factor analysis was used to extract utterance patterns for each genre, and three general purpose utterance patterns were clarified: question-answering type, attacking type, and proposal type. The results demonstrate the usefulness of utterance data construction and speech act analysis.