Abstract : This thesis is situated in the e-learning domain and in the context of the authoring of educative content (composed learning objects) by reuse of learning objects. This approach is supported by many models like SCORM or SIMBAD. Using such as approach by authors still difficult due to the complexity of the composition process influencing the quality of the content especially from a pedagogical point of view. We propose in our research works an author assistance approach with a set of analysis tools which offer best understanding of the true nature of the composed learning object and its degree of conformity to the author, the model and the use context requirements. Analyses are applied to the learning object's content, educative metadata and composition structure. The goal is to generate a learning object's detailed cartography. Automatic generated indicators reveal details about the different facets of the composed learning object especially the system view and the learner view. If the analysis is satisfactory additional metadata are then automatically calculated by the tools based on the reused learning objects' metadata. The learning object's composition can be guided by some conformity rules. Those rules can cover intended structural and semantic criterion. Our approach offers a support to promote the reuse of learning objects in an effective manner. It provides theoretical and practical elements allowing authors to control the "authoring by reuse" process. Consequently our approach allows authors to produce high quality composed learning objects.