Abstract : In this work, we study dynamic reconfiguration of collaborative communicating applications. Providing generic and scalable solutions for automated self-reconfiguration in group collaboration support systems can be driven by rule-based reconfiguration policies. To achieve this goal, we elaborate a dynamic graph-based modelling approach and we develop structural models that can represent the different interaction dependencies from different configurationrelated point of views: communication flows between the distributed machines, the networked deployment nodes, and the service composition. Our solution is based on graph grammars rewriting. We provide graph transformation to specify rules for changing deployment architecture while being in conformance to an architectural style. In order to handle the complex design of communicating collaborative system architectures and the related adaptation issues, we propose a multi-layer modelling approach. This approach assures generic solutions for automatic context aware adaptation. Our approach is based on the observation that semantic data analysis that can be exploited to manage priorities and more generally to manage communications. This allows us to represent, in a richer way, the semantics of the managed systems.