This paper presents the advantages of using a semantic network with default and exception abilities for a specific problem in language processing. More precisely, we show that the features of this semantic network provide an adequate representation of texts expressing a general rule and its exceptions. General statements are usually expressed in natural language by the presence of some specific words, and they occur frequently in our application domain. Our research is effectively integrated in a whole project which is an interactive user-aided tool for the description of requirements including the use of graphics and textual comments. The general cases of treatments are represented by semantic rules which are not strict and admit exceptions. The inferences supported by the network are justified by an interpretation in default logic.