Re: ShEx relation to SPIN/OWL

On 07/10/2014 09:12 AM, Karen Coyle wrote:
> On 7/10/14, 6:59 AM, Peter F. Patel-Schneider wrote:
>>> IMHO, it would be somewhat confusing to have two completely different
>>> semantics for OWL.
>> I don't think so.  You are using the same semantics by and large, but
>> instead of inferencing you are checking.
> This assumes that the "checking" rules and the intended OWL inferences are
> 100% compatible.

Well, yes, that does assume that in the cited work that the checking 
methodology does actually use the OWL semantics by and large, which I believe 
that it does.

> I believe there may also be the assumption that the ontology
> being "checked" is essentially private -- that there are no other uses or
> users of the properties whose context is different.

But what does this have to do with the semantics of the cited work?
> As someone from a community that shares widely (libraries, archives, museums),
> and where properties are used in many different contexts, the only axioms that
> should be attached to the ontology itself must be universal in nature. If the
> ontology constrains a property like "author" with an owl:maxCardinality for
> the purposes of checking, the effect of that on the data in the open web would
> be too restrictive for some uses within that community.

I think that we are talking about different things.  I was saying that the 
formal machinery of the cited work used very similar semantics to the OWL 
semantics, and thus that there would not be two completely different semantics 
for OWL.  You appear to be saying that the OWL ontology for a particular 
domain could be different from constraints that someone wants to put on data 
from that domain.  These are completely different.

> I think it comes down to the context within which you will be creating and
> sharing your data. If you operate in a closed or semi-closed environment, then
> OWL constraints may work for you. The LAM community instead is looking at
> development of "least ontological commitment" for sharing, with a sharable
> constraint language for those creating data and sharing in a more limited
> context.

And this can be achieved by having an OWL ontology for the domain plus 
separate constraints that are used in more limited contexts.


Received on Thursday, 10 July 2014 16:24:06 UTC