RE: Proposal to close ISSUE-12

Hi Peter!

>-----Original Message-----
>From: public-owl-wg-request@w3.org 
>[mailto:public-owl-wg-request@w3.org] On Behalf Of Peter F. 
>Patel-Schneider
>Sent: Wednesday, March 19, 2008 2:32 AM
>To: public-owl-wg@w3.org
>Subject: Proposal to close ISSUE-12
>
>
>ISSUE-12 points out that the mapping rules do not nicely handle most
>axioms that generate multiple triples.
>
>I propose to close this issue by modifying the mapping rules (and the
>reverse mapping) so that 
>1/ Annotations on axioms that generate single triples are as before 
>   e.g., ObjectPropertyDomain(Annotation(a "bar") r d) could become
>   	 _:x rdf:type owl11:Axiom
>	 _:x rdf:subject r
>	 _:x rdf:predicate rdfs:domain
>	 _:x rdf:object d
>	 _:x a "bar"
>2/ Annotations on axioms that generate a fresh blank node put the
>   annotation on that blank node, as is done already for negative
>   property assersions
>   e.g.,  DisjointClasses(Annotation(a "bar") c1 c2 c3) becomes
>   	  _:x rdf:type owl11:AllDisjointClasses
>	  _:x owl11:members SEQ(c1 c2 c3)
>	  _:x a "bar"
>3/ Other annotations on axioms that generate multiple triples (e.g.,
>   EquivalentObjectProperties) result in the triples being reified and
>   each annotation attached to each of the reified triples.
>
>peter
>
>
>

Point 3/ may produce a lot of duplication of information, in particular when
owl:RestrictionS are involved.

What about one approach to rule them all? For an axiom of the form

  s_1 p_1 o_1 .
  ...
  s_n p_n o_n .

and m different annotations "ap_k(v_k)" with annotation properties ap_k and
their values c_k, build an axiom annotation in the following way:

  _:x rdf:type owl11:Axiom
  _:x ap_k v_k , 1 <= k <= m
  _:x owl11:axiomStatement [ subject s_i ; predicate p_i ; object o_i ] , 1
<= i <= n

For single triple axioms, one /may/ decide to have a short form in the way
as it currently exist in the mapping WD. The advantage of such a special
treatment would be one triple less compared to the above encoding. So 5
instead of 6 triples if only a single annotation exists for the
single-triple axiom. This is not a large gain, IMHO, so I would rather opt
for having a unified treatment of /all/ annotated axioms in the way I
proposed above. 

And as I already wrote in an earlier answer to Alan, I would prefer to have

  * a dedicated shadow vocabulary for "subject", "predicate" and "object"
(avoiding the use of the RDF reification vocabulary)

  * having the original axiom still in the RDF graph, so that the annotation
construct exists in /addition/ to the encoding of the axiom.

The latter suggestion allows parsers to easily distinguish between
"relevant" code and pure comments, and it would not be necessary for them to
first reconstruct the original axioms. Also, it would not be necessary for
OWL-Full to specify special semantics for the annotation vocabulary, since
all relevant semantic information would already exist in the encoding of the
original axioms.

Cheers,
Michael

--
Dipl.-Inform. Michael Schneider
FZI Forschungszentrum Informatik Karlsruhe
Abtl. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: Michael.Schneider@fzi.de
Web  : http://www.fzi.de/ipe/eng/mitarbeiter.php?id=555

FZI Forschungszentrum Informatik an der Universität Karlsruhe
Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
Tel.: +49-721-9654-0, Fax: +49-721-9654-959
Stiftung des bürgerlichen Rechts
Az: 14-0563.1 Regierungspräsidium Karlsruhe
Vorstand: Rüdiger Dillmann, Michael Flor, Jivka Ovtcharova, Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus