Semantic Descriptions for Logical Content Generation

Human designers understand a range of potential purposes behind objects and configurations when creating content, which are only partially addressed in typical procedural content generation techniques. This paper describes our research into the provision and use of semantic information to guide logical solver-based content generation, in order to feasibly generate meaningful and valid content. Initial results show we can use answer set programming to generate basic roguelike dungeon layouts from a provided semantic knowledge base, and we intend to extend this to generate a range of other content types. By using semantic models as input for a content-agnostic generation system, we hope to provide more domain-general content generation.