Process modelling is a crucial phase of Business Process Management (BPM). Despite the many efforts made in producing process modelling tools, existing tools (languages) are not commonly accepted. They are mainly criticised for their inability to specify both the tasks making up the processes and their scheduling (their lifecycle models), the data they manipulate (their information models) and their organizational models. Process modelling in these languages often results in a single task graph; such a graph can quickly become difficult to read and maintain. Moreover, these languages are often too general (they have a very high expressiveness); this makes their application to specific types of processes complex: especially for administrative processes. In this paper, we present a new language for administrative processes modelling that allows designers to specify the lifecycle, information and organizational models of such processes using a mathematical tool based on a variant of attributed grammars. The approach imposed by the new language requires the designer to subdivide his process into scenarios, then to model each scenario individually using a simple task graph (an annotated tree) from which a grammatical model is further derived. At each moment then, the designer manipulates only a scenario of the studied process: this approach is more intuitive and modular; it allows to produce task graphs that are more refined and therefore, more readable and easier to maintain.