Different users (e.g. actual operators or other client software such tactical decision aids) may require / prefer different categorization representations. However, the Entity data model forces a choice of selecting a single option on the publisher.
Consequently, the logic for how to convert between different representations may become spread around different components of a system.
A better system architecture is to have knowledge and functionality with the producer of the data. That component can best judge the equivalence of different representations in the context of their own data.
This is a PSM issue; the DataSink interface has an EntityQuery filter that can be used to specify the Categorization variant in a PSM that actively services the request (i.e. isn't publish/subscribe). For a pub/sub PSM such as DDS it would be necessary to anticipate the Categorization demands in advance (when publishing), so that all variants are published.
Therefore, the needs of this issue can be satisfied by having the DDS PSM of the DataSink interface publish Entities ...