This study aims to clarify the composition of story content to help computers understand stories. In story content, the events peculiar to a genre occur intermittently. For example, school and sports festivals appear in school-themed genres. These events can trigger a story because they cause changes in the internal characteristics and relationships between characters, which in turn trigger the progress of the story. If computers can determine the events in a story, they will help understand its composition. Each story-event contains many strongly related words. For example, ``relay'' and ``runner'' appear in sports festival episodes. Therefore, investigating these tendencies is expected to contribute to the estimation of story-events. However, the amount of information obtained from comic texts is limited because they use illustrations and texts in a complementary manner. This makes it difficult for computers to obtain words from comics that characterize a story-event. To address this problem, we focused on the content similarities between comics and light novels. In this study, we estimated story-events in comics using the tendency of story-event words to appear in light novels. The results of this experiment indicate that we can calculate story-events in comics using a dictionary of story-events created by the proposed method.
17. ※1 T. Fujimura and H. Umemura, “Development and validation of a fa-cial expression database based on the dimensional and categorical model
of emotions.,” Cognition and Emotion, vol.32, pp.1663‒1670, 2018.
表情:恐怖,驚き,嫌悪(閉⼝)
<AIST顔表情データベース2017※1>
調査に⽤いるデータ
1つの表情種につき2名の⽇本⼈⼥性の画像を⽤いる
共感覚的表現:分類語彙表(国⽴国語研究所)や共感覚的表現に関
する論⽂・書籍から触覚に関連する共感覚的表現
(形容詞・オノマトペなど)36語を選定・抽出
14