Creating Character Networks from Comics Using Frames and Words in Balloons

Abstract

We create character networks from comics using the frames and the words in balloons to find and understand them. First, we build a dataset that contains a comic's characters and its words in balloons separated by frames. Second, we extract the semantic relations among characters using the words in the balloons and the frames. Third, we develop character networks by identifying the important characters with their relatedness and add semantic relations to the networks. We implemented our approach and successfully built character networks from four popular Japanese comics and evaluated our kinship relation extraction algorithms using them.