- From: Irene Polikoff <irene@topquadrant.com>
- Date: Thu, 19 Mar 2015 18:57:05 -0400
- To: Arthur Ryman <arthur.ryman@gmail.com>, "Peter F. Patel-Schneider" <pfpschneider@gmail.com>
- CC: Richard Cyganiak <richard@cyganiak.de>, "public-data-shapes-wg@w3.org" <public-data-shapes-wg@w3.org>
Arthur, I donıt think it should be SHACLıs job to unify "equivalent nodes". This would complicate things significantly and make everything muddy. SHACL would need to take the larger graph and transform it into some other graph. There would need to be rules defined for such transformation. This is complex to define and implement. I also donıt believe it is necessary. We seem to be in the agreement that OWL inferencing is outside of SHACLıs scope. Just as some external process takes graph G, applies OWL entailment and delivers graph H, some other external process could do the unification which ever way it wants. Irene On 3/19/15, 6:18 PM, "Arthur Ryman" <arthur.ryman@gmail.com> wrote: >On Thu, Mar 19, 2015 at 2:12 PM, Peter F. Patel-Schneider ><pfpschneider@gmail.com> wrote: >> This appears to be exactly backward. >> >> To handle entailments we have to go beyond the graph. >> >> peter > >Peter, > >Perhaps the following expresses this issue better... > >Given a graph G, entailment generally results in a larger graph H. >Simply counting triples in H may not give the desired results. >Overcounting may occur due to owl:sameAs and equivalent lexical forms >of literals. To get the desired result, we need to unify equivalent >nodes. Let E denote this equivalence relation on the nodes of H. Then >counting should be done on the quotient graph Q = H/E. > >-- Arthur >
Received on Thursday, 19 March 2015 22:57:51 UTC