Eric, I think your question is tool related... which means that the answer depends on the tool.
I don't think it's possible yet in NORMA, but my ActiveFacts project has a facility for handling it. I use some SBVR terminology; a model or schema is called a vocabulary. A vocabulary may import another vocabulary and define "correspondences" between the concepts (entity types, value types and aliases) of the two vocabularies. It's currently not possible to define alternate readings for imported fact types unless the fact type is objectified.
I introduced the idea of "alias" as being a local name for an imported concept, so an import may serve merely to translate from one language to another. Alternatively, if I have two vocabularies that overlap at some points, I can define the concepts that correspond. For example, one vocabulary might refer to a Client, another to a User, another to an Account, and each might have related support software that uses those terms. When I merge them into one software system, I can choose to declare that the terms correspond exactly to the same concept, and the resultant concept plays all roles as declared in each of the merged vocabularies.
That's the theory... I'm still working out how it'll function in practise. You can view the ORM diagrams of my metamodel by browsing the source code repository at http://activefacts.rubyforge.org/svn/examples/images/Metamodel. See Objects.png first, then Import.png. ActiveFacts is built on a base library of source code that's generated directly from this NORMA model.
The ActiveFacts home page is here