International Journal of Computational Linguistics Research Hajime Murai 8 3 2017 http://www.dline.info/jcl/fulltext/v8n3/jclv8n3_4.pdf Quantitative indicators of narrative analysis can be used to enhance the objectivity of story analysis. However, conventional narrative descriptions are not suitable for extracting the narrative functions of a story as a whole, nor have complex stories with parallel narrative structures been analyzed. With the ultimate goal of developing an automatic plot analysis method suitable for describing parallel storylines and punchlines, this paper proposes a prototype algorithm for automatic plot structure extraction focusing on agent vocabulary and behavior expressions. In the proposed prototype algorithm, categories and dictionaries for identifying agents and behaviors in story texts are constructed in Japanese and algorithms for extracting propositions and divisions of scenes are implemented based on the resulting agent and behavior vocabulary and expressions. Further, omitted agents are reproduced based on knowledge about agents and general rules of discourse. This enables the processing of fundamental data about story structure and the categorization of behaviors and scene divisions.