A complex system, integrates multiple sub-systems. (Carney, Fisher, & Place, 2005) Each sub-system contains some form of domain knowledge but the main difference that can be seen between a complex system and a set of simple system is that the knowledge the complex system presents is larger than the sum of the knowledge that each sub-system contains (Billaud, Daclin, & Chapurlat, 2015). In both cases, each sub-system is made of elements that are exploitable together, but two different sub-systems have information that can’t be used in concert. Therefore, the tacit knowledge, the knowledge contained in the interaction between the sub-system, is usually harder to extract than the knowledge contained in the sub-systems (Yahia, Lezoche, Aubry, & Panetto, 2011). In “simple cases”, the sub-system can be presented as a formal context, a cross-table of objects and their attributes, and the links between the diverse sub-systems can be represented as relational context, a cross-table containing the objects of two sub-systems, and modeling if two objects are in relation. In this paper, we aim to present the process to extract knowledge from such a model of complex system. We will show how to apply it to a use case of real data from the neurology domain.