Why not use labels (nip-32) for the type value? That would give you the ability to be more specific (with ontologies etc) but also remain more flexible going forward.
Overall I like the nip. I’ll try and add this label question directly to the nip shortly as well.