This paper proposes an approach for automatic know ledge extraction and matching that we apply to an aid system of management engineering diagnosi s called SADIM (Système d’Aide au Diagnostic d’Ingénierie de Management). SADIM aims to de tect the dysfunctions related to the enterprise management. It allows the knowledge acquisition from textual data related to the diagnosis, the matching and the assignment of witness sentences to the co rresponding key ideas. SADIM can also serve as a decision aid system which helps experts and socio- economic management consul tants to take decisions that would make enterprises reach the required standards through council interventions.