In this paper, we present the interpretation and generation processes of a spoken dialogue interface for ambient intelligence. The interface is automatically created for each specific environment and the interpretation and generation vary depending on the environment and its context. These processes rely on a dialogue tree structure. Several modules process the tree structure and the context information to produce specific dialogues for the current environment state. The interface has been implemented and evaluated in an ambient intelligence environment. Satisfactory objective and subjective evaluation results are shown at the end of the paper.