In most multiagent-based simulation (MABS) frameworks, a scheduler activates the agents who compute their context and decide the action to execute. This context computation by the agents is executed based on information about themselves, the other agents and the objects of the environment that are accessible to them. The issue here is the identification of the information subsets that are relevant for each agent. This process is time-consuming and is one of the barriers to increased use of MABS for large simulations. Moreover, this process is hidden in the agent behavior and no algorithm has been designed to decrease its cost. We propose a new context model where each subset of information identifying a context is formalized by a so called 'filter' and where the filters are clustered in ordered trees. Based on this context model, we also propose an algorithm to find efficiently for each agent their filters following their perceptible information. The agents receive perceptible information, execute our algorithm to know their context and decide which action to execute. Our algorithm is compared to a 'classic' one, where the context identification uses no special data structure. Promising results are presented and discussed.