JSAI2023

Presentation information

General Session

General Session » GS-10 AI application

[4T3-GS-10] AI application

Fri. Jun 9, 2023 2:00 PM - 3:40 PM Room T (Online)

座長:植野 研(東芝) [オンライン]

2:40 PM - 3:00 PM

[4T3-GS-10-03] Automatic generation of multiple endings based on analysis of endings of existing works

〇Shunsuke Kosaki1, Fumika Yoshii1, Jurin Sakamoto1, Hideyuki Nishimura1, Masato Irifune1, Hajime Murai1, Keiji Hirata1, Hironori Yoshida1, Kazushi Mukaiyama1, Emiko Tayanagi1 (1. Future University HAKODATE)

[[Online]]

Keywords:Automatic Ending Creation, Story Analysis, Story, Ending, Multi-Ending

Multi-endings, in which the ending changes depending on the choices the player makes in the story, are also widely used in computer games. However, there have not been many studies that address the structure of multiple endings from the viewpoint of narratology, nor have there been many studies on automatic generation of multiple endings from story scenarios. In this paper, we selected 51 existing works that are highly evaluated in the genre of otherworldly adventure tales and conducted a factor analysis of the endings in the works, extracting six types of patterns such as "an ending in which the character puts his/her own self first and returns home" and others. Using a large-scale language model, prompts were created according to the extracted patterns, and sentences were automatically generated as the basis of the scenario endings. The automatically generated sentences were then adjusted by humans and implemented in the game system.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password