Generation of Multi-Agent System Game Scenarios by The Semantic Score Method

Abstract

In this paper, we apply the semantic score method to a multi-agent system-based narrative generation system to generate game scenarios. The semantic score method is a method for analyzing the structure of a movie, in which scenes, which are semantic units of a story, are evaluated on a scale of complexity and resolution. By creating a semantic graph with the complexity of the evaluated semantic scores on the vertical axis and the scene number on the horizontal axis, the structure of a movie can be visually represented. We model game characters with roles and relationships as agents, and have them interact with each other in the simulation. The agent's behavior is scored for complexity and resolution, and the log data is output. The log data is given a semantic graph to verify how the scenario is output. Give a log data the characteristic semantic graph and verify how the scenario is output. As a result, we were able to generate a scenario that follows the development of the semantic graph by giving a characteristic semantic graph to the game scenario generation system using a multi-agent system.