W3C home > Mailing lists > Public > www-rdf-interest@w3.org > September 2002

Re: Can a property not have a uriref?

From: Sandro Hawke <sandro@w3.org>
Date: Wed, 18 Sep 2002 06:18:48 -0400
Message-Id: <200209181018.g8IAImX15006@wadimousa.hawke.org>
To: Zhang Jeff <jeffzhang726@yahoo.com.cn>
cc: www-rdf-interest@w3.org

[paragraphs reformatted to correct for mail user-agent mistakes]

Murray Spork (m.spork@qut.edu.au):
> > > > > I had assumed that all predicates must be named by some
> > > > > uriref - this seems obvious where a predicate appears in a
> > > > > simple s o p triple.  ...  But today I thought of another
> > > > > possible counter example (where we are dealing with
> > > > > reification) that may actually make sense in some
> > > > > circumstances.

Sandro Hawke (sandro@w3.org):
> > > > Actually, there are lots of times when it makes sense to have
> > > > arcs without URIRef labels, even without reification.  Two
> > > > that come to mind are Currying and using properties when you
> > > > only know their unambiguous properties.

Zhang Jeff (jeffzhang726@yahoo.com.cn):
> > > If you really want it, why not use an instance of
> > > "http://www.w3.org/...#Property" which is just
> > > a property without any specific meaning?
> > > OR introduce an "anonymous property",into RDF
> > > vocabulary?
> > > OR in application level,such as
> > > "http://www.mycompany.com/unknowProp"  typed as
> > > rdf:Property ? It is easy to deal it as a "general
> > > property" in application level.
> > > (I don't have much knowledge of AI or theories of
> > > models or math and I just can not see the
> > > complexity
> > > of this problem. Is it really so difficult? )
> > > And,is there any similar concepts in logics or
> > > relation algebra or other knowledge
> > representations?

Sandro Hawke (sandro@w3.org):
> > You're 99% right.  Skolemizing an RDF graph (labeling unlabeled
> > nodes and arcs with UUIDS) does not change its meaning as an
> > assertion.  The other 1% comes from not wanting UUIDs cluttering
> > up databases, and wanting to use RDF with non-assertional
> > attitudes, such as to discuss what someone else said/signed in
> > RDF.  (I say "Something is red."  If you put "uuid:wer4353535 is
> > red" into your DB the results will generally be the same.  But you
> > shouldn't claim I said uuid:wer4353535 was red.)

Zhang Jeff (jeffzhang726@yahoo.com.cn):
> I see.  If we introduce a general resource that standing for just
> "some thing" and a general property for "some property" into the
> vocabulary, somewhat like wildcards,can the problem be resolved?  So
> you can say" rdf:something is red" or "myNameSpace:something is
> red".  There are concepts of indefinite pronouns in natural
> languages and I think it is natural to introduce similar things into
> artificial languages like RDF.

If we just had one rdf:something, we couldn't say: "something is red
and large, and a different something is green and small."

There is a general mechanism for this in RDF; it's called unlabeled
nodes or bNodes; it gives you as many "somethings" as you want.  But
its not allowed for predicates/properties.  So you can say

   Tom likes Jane
or
   something likes Jane
or
   Tom likes something
but not
   Tom has-some-relationship-to Jane

I think this is a silly restriction, but as you pointed out, it's not
very different from 
   Tom uuid:b546a35a-caef-11d6-87f4-0050ba4812a6 Jane
where we don't know what "uuid:b546a35a-caef-11d6-87f4-0050ba4812a6"
means. 

   -- sandro
Received on Wednesday, 18 September 2002 06:20:10 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:56 GMT