Re: Can Shapes always be Classes?

On 11/20/2014 2:30, Peter F. Patel-Schneider wrote:
>
>
> On 11/04/2014 09:16 PM, Holger Knublauch wrote:
>> I believe there is a fundamental difference in how the various 
>> proposals treat
>> the relationship between resources and their shapes:
>>
>> - In OWL and SPIN, constraints are attached to classes. rdf:type 
>> triples are
>> used to determine which constraints need to be evaluated for a given 
>> instance.
>
> This is false for both OWL constraints and Stardog ICV.

Yes I was simplifying a bit. You are right: some constraints are 
attached to properties (owl:FunctionalProperty, domain and range etc).

>
> OWL constraints utilize OWL axioms as constraints.  Although many OWL 
> axioms are used to define classes not all are.  Although most OWL 
> axioms are concerned with class membership not all are.  Many OWL 
> axioms are unconcerned with explicit rdf:type triples.
>
> Stardog ICV constraints are also not all like class definitions. For 
> example, Stardog ICV can be used to check domains and ranges for 
> properties or subproperty relationships.  Stardog ICV can also have 
> constraints that do not depend on explicit rdf:type properties in the 
> input.  See the http://docs.stardog.com/icv/ for examples of all of 
> these.  (You would have to simplify an example to get the last aspect 
> of Stardog.)

All I wanted to say was that the starting point to determine which 
constraints to evaluate is usually the type of an instance (or of a 
property). I am fully aware that there are exceptions, also including 
global constraints in SPIN. But all information needed to determine how 
to evaluate constraints can be derived from the graph itself, and 
doesn't require external input such as a "starting point".

>
>> - In the original Resource Shapes and ShEx, Shapes are stand-alone 
>> entities
>> that may or may not be associated with a class. Other mechanisms than 
>> rdf:type
>> are used to point from instances to their shapes.
>
> Resource Shapes and ShEx appear to work by performing type recognition 
> and then type checking.   That is, they determine which objects belong 
> to particular types and then validate by checking that objects belong 
> to the right types.  They can do this without explicit rdf:type 
> triples in their input, but they are, in effect, computing rdf:type 
> triples and then checking for their existence on particular objects.
>
> This kind of processing can be easily handled in OWL constraints.   
> Many kinds of recognition constraints can also be handled in Stardog ICV.

We all agree that OWL can cover some scenarios. For OWL to be considered 
a serious starting point, someone should walk through the User Stories 
and work out whether and how OWL could be used. I believe OWL is too 
limiting overall, so I guess it could only be used as one vocabulary 
among others. Some solution to leverage existing OWL ontologies should 
certainly be a goal of this WG.

Holger

Received on Wednesday, 19 November 2014 22:54:07 UTC