Proper names often constitute a problem in translation. This contribution deals with an ontology which represents the basis for a multilingual database of proper names, Prolexbase. It is being set up for treatment of proper names in the framework of the Prolex project, a research programme supported by the French Ministry of Industry in collaboration with two firms working on the market of language technologies: Systran and Exalead. The aim of this collaboration is to create a multilingual database of proper names containing information applicable to machine translation, computer aided translation, data research as well as spelling dictionaries. These particular aims guided the creation of the ontology whose description will follow. Beside a set of language-dependent and language-independent relations associated with a logical model, the data- base is founded on a four level ontology: the level of instances (the proper names such as they appear in a written text in a specific language), the linguistic level (the level of so called “prolexemes”), the conceptual level (the numerical pivots) as well as the metaconceptual level (types and supertypes). We will describe here the different levels of the ontology and their implementation in the database using French and German examples.