Prevent injection of inappropriate annotation property declarations

[cmungall]

not sure if this belongs on obo tracker, as a robot validation check, here, somewhere else

many ontologies declare rdf:type to be an AnnotationProperty:

SELECT ?p ?y ?g WHERE {GRAPH ?g {<http://www.w3.org/1999/02/22-rdf-syntax-ns#type> ?p ?y}}

rdf:type,owl:AnnotationProperty,obomerged:OSCI
rdf:type,owl:AnnotationProperty,obomerged:CLO
rdf:type,owl:AnnotationProperty,obomerged:ICDC
rdf:type,owl:AnnotationProperty,obomerged:CIDO
rdf:type,owl:AnnotationProperty,obomerged:SWO
rdf:type,owl:AnnotationProperty,obomerged:OPMI
rdf:type,owl:AnnotationProperty,obomerged:OLAM
rdf:type,owl:AnnotationProperty,obomerged:HTN
rdf:type,owl:AnnotationProperty,obomerged:PSDO
rdf:type,owl:AnnotationProperty,obomerged:OHMI
rdf:type,owl:AnnotationProperty,obomerged:LABO
rdf:type,owl:AnnotationProperty,obomerged:OAE
rdf:type,owl:AnnotationProperty,obomerged:FBBT
rdf:type,owl:AnnotationProperty,obomerged:STATO
rdf:type,owl:AnnotationProperty,obomerged:OBIB
rdf:type,owl:AnnotationProperty,obomerged:ORNASEQ
rdf:type,owl:AnnotationProperty,http://purl.unep.org/sdg/merged/SDGIO
rdf:type,owl:AnnotationProperty,obomerged:OHPI
rdf:type,owl:AnnotationProperty,obomerged:LTHIDO
rdf:type,owl:AnnotationProperty,obomerged:TSO
rdf:type,owl:AnnotationProperty,obomerged:CTO
rdf:type,owl:AnnotationProperty,obomerged:ONS
rdf:type,owl:AnnotationProperty,obomerged:KTAO
rdf:type,owl:AnnotationProperty,obomerged:VIO
rdf:type,owl:AnnotationProperty,obomerged:ICEO
rdf:type,owl:AnnotationProperty,obomerged:VIO-ImmPort

This is likely propagated from a single source

I can't point you to a specific W3C spec that says this is incorrect, it's not really punning since rdf:type is not a property. But this could have unanticipated effects and is objectively confusing and I assume unintended.

This is partly related to OBOFoundry/OBOFoundry.github.io#1443, although here we are not injecting into an OBO, in fact we are not even injecting into a vocabulary since rdf:type doesn't exist as an entity at the OWL level.

It seems OMO is a good place to have a rigorous enforced data model for annotation properties that is restrictive enough to forbid this kind of thing.

[matentzn]

Note that we recently added a robot check to cover that particular issue https://github.com/ontodev/robot/pull/867/files

But this does not change the general need for a better general checks.

In ontologies like CL and Mondo, we now employ a strategy for limiting the number of legal annotations using sparql - but these are tailored specifically to these ontologies.