Re: Use of owl:equivalentProperty with owl:AnnotationProperty

On Jan 17, 2017 11:59 AM, "Chris Mungall" <cjmungall@lbl.gov> wrote:


AFAICT you're allowed to pun between an annotation property and individual
https://www.w3.org/TR/owl2-new-features/#F12:_Punning

So you can make a sameIndividuals axiom about the two IRIs.

In the RDF / OWL full interpretation, they two properties would be the same.


If we are ignoring the OWL 2 DL conditions, then it is possible to use an
rdf equivalentProperty to two annotation properties, which under the OWL
RDF semantics would make their extensions the same.

However it would not be possible to parse the resulting file into OWL 2
Axioms (something true in theory for non DL OWL, but expecially
 problematic in this case).

Under OWL RDF semantics the equivalentProperty approach would work for
annotation assertions, but not for axiom annotations, where the annotation
 property is mentioned rather than used.

owl:sameAs might work, but is probably incorrect, since it asserts
intensional equivalence, which is unlikely when aligning vocabularies.

Asserting P ⊑ Q and Q ⊑ P would fit most use cases, and could generate
appropriate results for many use cases. Since the annotations are generally
"ignore", it could be relatively cheap to add a new Axiom as syntactic
sugar. The result would not be standard OWL 2.

It also be relatively easy to add yet another special case to the OWLAPI
RDF  triple consumer that would translate the troublesome triple into two
subproperty axioms. The two generated axioms could be linked by annotation.
Everyone loves the OWL RDF Consumer.

It would be technically  possible to perform the reverse transformation to
equivalentProperty on output to RDF, but that would require generating
dubious output, and would only work for RDF formats.

But I suspect this isn't really required. For mapping between vocabularies
you can just roll your own annotation property for saying two things match,
or use the OWL version of SKOS?


I'm not sure that the SKOS matching properties are appropriate here; they
would relate the "aboutness"  of the concepts of the properties rather than
the properties themselves*.
(Of course, SKOS itself is quite problematic OWL; a relevant example here
is the use of annotation properties for most, but not all labels; labels
are a fundamental part of the domain of discourse for Knowledge
Organization Systems, and semantic conditions that could otherwise have
been expressed are relegated to comments.)

Simon
* language should not be taken to imply commitments beyond Intentional
stance.