I'm not in a position to comment on any recent Doctool versions as it has been close to a decade since I touched a new version, but I am familiar with the approach. There is certainly a stronger emphasis on sample population as the whole system focuses on the lexical level. The system does not use ORM object types at all. The primary display is called a 'fact type diagram', with a single fact type diagram representing multiple elementary fact types in ORM.
Given the lack of object types, which are fundamental for forming the ORM verbalization patterns, the Doctool verbalization is generally limited to inserting sample data into the sentence patterns written by the user. This is a much different style of verbalization and validation than we use in the NORMA tool, which forms logical statements based solely on the model, not a population. Both approaches have benefits.
Karl, if you've used the tool, please let the forum know if I've misrepresented anything here.
(Back to NORMA)
To validate a sample population against all constraints effectively requires a full physical implementation of the constraints in the system. This would also likely be a different implementation than a production system with soft (deontic) constraints so that invalid data could be entered. If this happens in the tool, it will likely be based on such a mapping of the system, with data entry still done at the conceptual (fact type and object type) level.