[2N1-GS-6x] Language media processing

9:00 AM - 9:15 AM JST(12:00 AM - 12:15 AM UTC)

[2N1-GS-6x-01] Prototype development of automatic sequel generation function based on changed attributes in existing sequel stories

〇HAJIME MURAI¹ (1. Future University Hakodate)

[[online]]

Keywords:

Story,Sequel,Automatic generation

Large-scale language models are rapidly developing, but various challenges remain when it comes to automatically generating stories with complex structures, such as full-length stories. In this study, the structure of full-length stories is focused on, which are composed of multiple stories in succession. First, in order to clarify the changes in attributes in sequels to existing works, the differences in the attributes of approximately 200 pairs of prequels and sequels were classified into two levels, consisting of 29 items in 9 categories. The characteristics of the attribute changes were extracted using methods such as factor analysis, along with changes in the protagonist's main objectives. Then the extracted characteristics were used to create a prototype program that generates a plot synopsis for a sequel that corresponds to the objective of the prequel, in line with the changes in the attributes of the existing work.