Host: The Japanese Society for Artificial Intelligence
Name : 34th Annual Conference, 2020
Number : 34
Location : Online
Date : June 09, 2020 - June 12, 2020
This paper reports on the extension of an automatic narrative generation game developed by the authors in a previous study. The first extension is aimed at expanding the knowledge bases, and the second extension is focused on evaluating the generation result. The automatic narrative generation game consists of a mechanism that prepares the framework of the story and another mechanism that generates fragmentary stories. The game generates stories by inserting fragmentary stories in the framework. In this study, we attempted to generate a story using existing novel and folk tale texts as the knowledge for story generation. During the generation, we employed a quantitative evaluation using Doc2Vec’s method and considered a method of generating a story that is not as similar as possible to the original story. Furthermore, the generation results are evaluated via a method that refers to a generative adversarial network, and natural stories and unnatural stories are generated.