In this article, we trace the genesis of the French derivational database Démonette and show how its architecture and content stem from recent theoretical developments in derivational morphology and from user needs. The development of this large-scale resource began a year ago as part of the Demonext project. Its conception is grounded in a theoretical approach where the lexemes are connected by derivational relations within derivational families which in turn fit into paradigms. More precisely, Démonette is a partial implementation of ParaDis, a paradigmatic model of morphological representation designed for the description of regular processes and of form-meaning discrepancies. The article focuses on the principles that govern the morphological, structural and semantic encoding of morphologically complex lexemes in Démonette and illustrates the range of form-meaning discrepancies with a variety of examples of non-canonical word formations.