Scene Situation Identification by Game Scenario Data

Abstract

In recent years, the rapid development of machine learning has been actively applied to understanding creative works. Among the various creative works, game scenarios have features that are not found in novels and manga, which are other creative works with stories, such as the fact that the characters' lines change depending on the player's choices. In that case, the characters' emotions are clearly described in the scenario as motions. These features suggest that the game scenarios can be used to understand stories and other creative works by artificial intelligence. In this study, we solved the problem of identifying geographic situations from conversations to confirm the usefulness of game scenarios. In places where the actions that can be taken in general are limited, the similarity in the content of conversations and the words that appear in them resulted in high accuracy. We confirmed the usefulness of game scenarios for understanding geographical situations.