Semantic labelling is the assignment of meaning-based labels — subject, entity type, business domain, regulatory category and taxonomy terms — to data assets and their elements, giving policy something more expressive than a sensitivity level to act on.
What is semantic labelling?
Sensitivity tells you how protected something is; semantic labels tell you what it is about. Policy usually needs both — "restricted" is a weaker instruction than "restricted, clinical, subject to the same purpose limitation as the source consent". Semantic labels also make policy portable across systems, because they describe meaning rather than the schema or folder structure of one source. Generating them automatically is what allows classification to keep pace with the volume of unstructured data that AI consumes.