Semantic Content Generation Framework for Game Worlds

As virtual worlds in games become larger and more detailed, the need for rich, interactive content to realistically populate these worlds becomes greater. Content is expensive and slow to create manually and does not scale once created. Procedural content generation offers an attractive alternative for providing this content. There are several methods for generating different types of content including terrain and some types of organisms. However, these methods are hard to control and can give inconsistent results. A method of providing a high level, semantic context to such procedural methods may offer a potential solution. Such an approach may also allow manually created content to be placed appropriately and in a scalable manner in the world. Finally, semantic knowledge may be used to annotate content to allow greater interactivity and improved AI. This paper suggests such a method. Making use of semantic networks for storing knowledge about the potential content of the virtual world and using stateful graph traversal algorithms to convert the semantic knowledge into concrete instances, this method supports the procedural generation of rich complex content for virtual worlds.