The setting-up of a database from which a corpus and associated information (whether syntactic or semantic, etc.) are derived is not a natural undertaking in non-computational linguistics. This article sets out to present how such a technique can be exploited within the context of a research project focussing on the French preposition dans dans.