A study of automatic visual styling system for digital books

Abstract

In the advancement of the digital book technology, the desire that we wants to read books more happily and convenient has risen. Actually, the digital book reader that we can freely specify the background color etc. appears. However, a current visual style values only the easiness of visibility, and doesn't consider the content of the novel. They do not help us to understand the contents. In the present study, we propose the automatic visual styling system used for the digital book. The system performs (1)Scene division, (2)Clustering of similar scene and (3)Mood presumption of each cluster. The result is used for visual style of novel. In this paper, we evaluated the first (1) Scene division based on time, place, and person information. By the experiment, the result of F value 0.216(0.413 correct answer in one sentence before and after) was obtained.