Detection of the helper types from story in multimedia

Abstract

A story unfolds through the relationships formed by the major and minor characters. The major characters are central to the main plot of the story, creating and resolving the conflicts. In contrast, the minor characters tend to play the role of awakening the major characters or solving their problems. Although the minor characters appear less frequently than the major characters, they play crucial roles in that they create tension, provide clues that lead to a solution, and so on, thereby making the story more interesting and engaging. Therefore, an analysis of the minor characters is essential for analyzing the entire story. However, the existing character analysis methods such as Character-net and RoleNet are not entirely suitable for analyzing the roles of minor characters in a story, even though they are adequate for classifying whether a character is one of the major characters or a minor character based on an analysis of the cumulative outcomes of the story. For an accurate analysis of the roles of minor characters, there is a need to study the pattern in which the characters appear as the story progresses, not to analyze the story based on cumulative outcomes. Accordingly, this paper proposes a method of classifying minor characters as either a mentor or a best friend, based on a three-dimensional visualization of Character-net which can depict the story progress. In order to make such distinctions, a set of classification rules based on appearance patterns and density was proposed, and forty characters chosen from twenty-five movies were examined for the purpose of performance evaluation. The results showed that the proposed method had a significant classification performance with an F1-measure of 0.565 with respect to the forty characters.