Design of game worlds is becoming more and more labor-intensive because of the increasing demand and complexity of content. This is being partially addressed by developing semi-automated procedural techniques that help generate (parts of) game worlds (e.g terrains, cities and buildings). However, most level editors rather deficiently capture and deploy designer's intent. For example, common positional or functional relationships between objects are usually limited to pre-processing a number of anticipated cases. In this paper we propose a novel scheme for specifying high-level semantics of objects within a game virtual world, and in particular we illustrate its application to a variety of layout solving problems raised by procedural generation methods. Our approach combines the genericity of a semantic class library with the power of a layout solver, and it shows to be both very flexible and effective. Moreover, this scheme can be useful for improving both manual, automated and mixed modeling techniques, always leading to a more efficient layouting process for game worlds. We conclude that by allowing designers to capture more of their intent and real-life knowledge in the objects with which they populate a game world, the integration of semantics will strongly contribute to stimulate content reusability, enrich the game play, and eventually also significantly cut down design duration and cost.