We provide a formal characterization of a notion of contextual taxonomy, that is to say, a taxonomy holding only with respect to a specific context. To this aim, a new proposal for dealing with “contexts as abstract mathematical entities” is set forth, which is geared toward solving some problems arising in the area of normative system specifications for modeling multi-agent systems. Contexts are interpreted as sets of description logic models for different languages, and a number of operations on contexts are defined. Using this framework, a simple scenario taken from the legal domain is modeled, and a formal account of the so called open-texture of legal terms is provided characterizing the notions of “core” and “penumbra” of the meaning of a concept.