There are currently no well-established techniques or tools to build adaptive 3D user interfaces. The aim of this research is to overcome identified shortcomings by providing a model and framework for the semantic enrichment of scene graphs for task based adaptation in 3D user interfaces. The semantic enrichment of scenes can play an extremely important role in enabling the viewers to understand and interact with the usually complex and incomprehensible visualized information, in simple, intuitive and user-friendly ways. The goal of our research is to develop a general model and framework supporting the adaptations of 3D user interfaces. The general model and framework relies on the formalization of the vertical correspondence between two layers: the spatial layer (the graphics) and the semantic layer (the concepts).