This article presents the preliminary steps to the implementation of detection and correction strategies for the erroneous use of N+N structures in the written productions of French-speaking advanced users of English. This research is carried out as part of the grammar checking project "CorrecTools", in which errors are detected and corrected using linguistic-based NLP techniques. We use information from comprehensive and student grammars as well as the results of linguistic research in order to identify a set of constraints on the formation of N+N structures. We use these constraints to propose a classification of the N+N errors found in our corpus and design detection and correction strategies for each error type. (Contains 1 table.)