RE: Mapping to RDF Graphs and reification

Jeff Thompson wrote:

>Thanks everyone for the comments.  Note that the message from Tim I
>quoted was
>about provenance.
>http://lists.w3.org/Archives/Public/semantic-web/2007Jan/0088.html
>This issue is confusing because a first instinct for annotations is to
>use them
>to give the provenance of a triple.  Indeed the first annotation example
>in
>the OWL 2 primer appears to be addressing this:
>
>Individual f:John
>   Facts: Annotations: dc:author Individual(f:peter)
>                       dc:creationDate "2008-01-10"^^xsd:date
>                       rdfs:comment "A simple fact about John"
>   f:hasWife f:Mary
>
>The confusion is that this is arguably NOT provenance data about who is
>being quoted
>to say that f:John f:hasWife f:Mary, but rather an assertion that peter
>put this
>triple into THIS ontology according to the semantics that peter knew
>that
>this ontology has for f:John, etc.  This is OK, and within the intended
>semantics
>of OWL 2 annotations.

Actually, the intended semantics of OWL 2 annotations is that there should not be semantic meaning for them at all. Whether they are used to represent provenance data or something else is up to the ontology's author, and the interpretation of annotations is up to the ontology's users, including specialized processing tools. OWL 2 is completely ignorant about what an annotations such as

  dc:creationDate "2008-01-10"^^xsd:date

is intended to mean. 

In this respect, OWL 2 isn't very different from OWL 1. In OWL 1, there have already been annotations, but only for URIs/Entities. For example, it was possible to write:

  ex:peterparker dc:creationDate "2008-01-10"^^xsd:date

What is new is that OWL 2 now also allows for annotating /axioms/ in an ontology, not only URIs. But in any case, OWL 2 has no idea about the intended meaning of an annotation property (such as dc:creationDate), and it also doesn't understand what the value of an annotation is intended to mean.

>If however there is someone else, such as Bob, out there in the world
>who is the
>provenance source and is being quoted as saying that John hasWife Mary,
>then it
>matters to quote how Bob said this.  If the ontology includes f:Mary
>sameAs f:SecretAgent99,
>then Bob may never have said that John hasWife SecretAgent99, so it is
>not correct
>to use an OWL 2 style annotation to *quote* what Bob originally said,
>because the
>semantics of OWL 2 annotations absorb all the sameAs and other
>inferences in the ontology.

Now, there is a distinction between OWL 2 DL and OWL 2 Full. OWL 2 DL completely ignores annotations when it comes to semantic inference. So if you have

  ex:peterparker dc:creationDate "2008-01-10"^^xsd:date

and

  ex:peterparker owl:sameAs ex:spiderman

OWL 2 DL does /not/ give you

  ex:spiderman dc:creationDate "2008-01-10"^^xsd:date

However, in OWL 2 Full, which builds on top of the RDF Semantics, you really get this inference. So in OWL 2 Full, unlike OWL 2 DL, an annotation is actually an assertion about the resource denoted by an URI, instead of a comment assigned to that URI. In our example, the annotation is not on the URI ex:peterparker, but it is some assertion about the denoted person, Peter Parker, who happens to be alsoe denoted by the URI ex:spiderman. 

But note that it is completely unclear what this assertion is about. OWL 2 Full, just as OWL 2 DL, completely ignores the intended meaning of annotation properties and their values. In our example, OWL 2 Full does not even know about the annotation(?) property 'dc:creationDate'. That this is a property at all only becomes clear (i.e. follows from the RDF semantics) from its occurrence as a predicate in the above example triple.

>The problem is that many people will see OWL 2 annotations and leap on
>them to solve
>the desperate need for provenance data in RDF/OWL, but they shouldn't.
>
>So to my question: Is it worth my time to try to convince public-owl-
>comments@w3.org
>that some words of warning should be added to the specification?

Perhaps, it makes some sense to have a comment in the Primer and/or in the RDF-Based Semantics (OWL 2 Full) on the technical difference between OWL 2 DL and OWL 2 Full I pointed to above. Some clarification on this might be of value in order to avoid confusion. So feel free to speak up!

>- Jeff

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