Our societies increasingly depend on huge amounts of textual data stored in large repositories, e.g. on the Web, in government archives, etc. Developing ways of bringing order to, and handling, this unstructured information is thus a major challenge of our times. Some of the most promising approaches for enabling computers to represent the required conceptual information appear to be ontology based ones. It is the precise path taken by this project, in application to lexical databases and WordNet in particular.The number of applications of WordNet (in natural language processing, knowledge management, and information systems) is rapidly growing despite its well-known defects. Among these flaws, the project focuses on the particular issue of nominal systematic polysemy. The strongly hierarchical structure of - WordNet, as of most lexical ontologies, cannot presently cope with this kind of ambiguity, as it involves orthogonal non-taxonomic relations between categories. As a result, systematic polysemy is treated in a rather ad hoc way in WordNet, compensating for missing orthogonal relations either by enumerating meanings as if disconnected, or by introducing problematic multiple inheritence. This lack of methodical treatment produces notorious incoherences and compromises the results of applications using WordNet.The aim of this fellowship is to address these issues in an ontology driven approach, showing that systematic polysemy results from orthogonal dependence relations between the denoted entities. Drawing on philosophy and computational linguistics and ontology, our objective is to map patterns of nominal systematic polysemy into patterns of dependence.