Generating Character Lines in Four-Panel Manga using Progress and Line Samples

Abstract

Automatic content generation based on natural language processing is an active research area, especially for story generation. Research on story generation has focused on generating consistent text pertaining to characters' actions and events; however, there have been few studies on generating characters' lines (utterances) and dialogues. This paper proposes a line generation model that bridges the gap between automatic story generation and practical content generation. Proposed model estimates the progress of a given plot from context and uses line samples to generate characters' lines based on the plot. Line-generation experiments using a novel dataset created using four-panel mangas revealed that our model can generate lines that follow given plots and match characters' personalities.