On 6/10/2017 18:44, Bohms, H.M. (Michel) wrote:
See https://twitter.com/wohnjalker/status/915982539747028992
<https://twitter.com/wohnjalker/status/915982539747028992>
ĆGreat summary! š
As a slightly more serious response, I agree that URIs from the OWL
namespace may be useful even without OWL semantics. owl:imports is
clearly useful, and even referenced by the SHACL spec. owl:versionInfo
and the deprecation mechanisms can be useful, but they don't carry OWL
semantics. Whether owl:DatatypeProperty and owl:ObjectProperty provide
value is a matter of debate. I believe as long as there are sh:class
and sh:datatype or sh:nodeKind constraints in place, then there is no
need for them. I am not fond of global property axioms in general, but
that's another topic.
Maybe there is value in going through the ways that people have used
OWL so far and verify how many of them were really designed for OWL
(DL) inferencing. Maybe you have examples of axioms in your world,
that you could share here so that we can see what would be left that
isn't covered by SHACL or other non-OWL vocabularies.
>well so far the distinction between attributes/datatypeproperties and
relationships/objectproperties has proven useful since they in the end
say something about intrinsic properties of things and the more
role-based extrinsic properties towards other independent things which
is a quite basic notion in conceptual modelling not only in LD/SW but
in any other earlier modelling system. But always interesting of
course to rethinkā¦.
> owl:import hopefully obsolete in a future where all is dereferenceableā¦
>owl:equivalent could be two way rdfs:subClassOf of course
>owl:sameAs is seen as important but has big issues too (is it really sameAs that you want etc. ie
what does it mean, donāt you really want a weaker thing; and in CWA
can be done via UNA anyway)
> unionOf/intersectionOf but I expect they have counterpart in shacl
> inverse properties
>disjointWith /propertydisjointWith
> So actually main concern is distinction in attributes and relationships (or if you like values
and references) in the endā¦. š
On the owl:Datatype/ObjectProperty topic, why would
sh:datatype/sh:class/sh:nodeKind not be sufficient? The only difference
is that the latter are per-shape, which is similar to how most other
languages such as UML or XML handle these things - the concept of global
property declarations are pretty unique to semantic web technology. But
for a tool there is little difference on whether you check (in SPARQL)
?property a owl:DatatypeProperty
vs
?ps sh:path ?property .
?ps sh:nodeKind sh:Literal .
Note that even the former case (in OWL) is not sufficient - you would
still need to add code to query for the cases of untyped properties or
properties that have rdf:type rdf:Property and then declare an
rdfs:range or owl:allValuesFrom restrictions. The SHACL variant looks
more pragmatic compared to that.
Holger
--
You received this message because you are subscribed to the Google Groups "TopBraid
Suite Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
