W3C home > Mailing lists > Public > public-owl-wg@w3.org > November 2007

RE: ISSUE-3: REPORTED: Lack of anonymous individuals

From: Boris Motik <boris.motik@comlab.ox.ac.uk>
Date: Fri, 9 Nov 2007 11:15:04 -0000
To: "'Giorgos Stoilos'" <gstoil@image.ece.ntua.gr>
Cc: <public-owl-wg@w3.org>
Message-ID: <001201c822c1$c6f4e5b0$2711a8c0@wolf>

You can't say this in the abstract syntax: the OWL 1.0 DL abstract syntax has been restricted to allow only for tree-like
connections between individuals.

With my examples I just wanted to make these two points:

- We can't really give up on tree-like connections without sacrificing decidability of ontology entailment (which gives us a
negative answer to ISSUE-46).

- If we give up on "true" anonymous individuals altogether and treat them just as Skolem constants, then we can allow for arbitrary
connections between individuals (which gives us a positive answer to ISSUE-46). In this case, we would, of course, need a way to
state this in the OWL 1.1 Functional-Style Syntax. Such an extension, however, would be trivial.

	Boris

> -----Original Message-----
> From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org] On Behalf Of Giorgos Stoilos
> Sent: 09 November 2007 11:08
> To: 'Boris Motik'
> Cc: public-owl-wg@w3.org
> Subject: RE: ISSUE-3: REPORTED: Lack of anonymous individuals
> 
> 
> Hi,
> 
> Indeed if you use triples syntax or FOL syntax (as you did initially) or any
> other syntax that allows you to play with variable, then you can construct
> these things/problems. But I was asking about using OWL abstract syntax.
> 
> Greetings,
> -gstoil
> 
> > -----Original Message-----
> > From: Boris Motik [mailto:boris.motik@comlab.ox.ac.uk]
> > Sent: Friday, November 09, 2007 12:50 PM
> > To: 'Giorgos Stoilos'
> > Cc: public-owl-wg@w3.org
> > Subject: RE: ISSUE-3: REPORTED: Lack of anonymous individuals
> >
> > Hello,
> >
> > You can use bnodes in RDF data arbitrarily. Take a look at
> >
> > http://www.w3.org/TR/rdf-primer/#structuredproperties
> >
> > There, you'll see triples containing identifies form _:xxx; all of these
> > are bnodes. Each such identifier is taken to represent one
> > existentially quantified variables.
> >
> >
> >
> > The bnodes in RDF are the same as labelled nulls in databases. Database
> > people have studied in depth what kind of semantics is
> > appropriate for null values. A really good paper on this topic is the
> > following:
> >
> > Tomasz Imielinski, Witold Lipski Jr.: Incomplete Information in Relational
> > Databases. J. ACM 31(4): 761-791 (1984)
> >
> > The practical consequences, however, are rather severe: answering queries
> > with labelled nulls is NP-complete. This is one of the
> > main reasons why practical database systems don't implement labelled
> > nulls. (Another reason is that your answers are not so much
> > better even if you use labelled nulls.)
> >
> > Boris
> >
> >
> > > -----Original Message-----
> > > From: Giorgos Stoilos [mailto:gstoil@image.ece.ntua.gr]
> > > Sent: 09 November 2007 10:18
> > > To: 'Boris Motik'
> > > Cc: public-owl-wg@w3.org
> > > Subject: RE: ISSUE-3: REPORTED: Lack of anonymous individuals
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: Boris Motik [mailto:boris.motik@comlab.ox.ac.uk]
> > > > Sent: Thursday, November 08, 2007 2:14 PM
> > > > To: gstoil@image.ece.ntua.gr; public-owl-wg@w3.org; 'Carsten Lutz'
> > > > Subject: RE: ISSUE-3: REPORTED: Lack of anonymous individuals
> > > >
> > > > Hello,
> > > >
> > > > Here is an explanation how anonymous individuals in ABoxes correspond
> > to
> > > > conjunctive queries. I will use a "pidgin" LaTeX
> > > > first-order logic notation for this. I'll use _:x for anonymous
> > > > individuals, and I'll use != for inequality (DifferentFrom)
> > > > assertions and & for conjunction.
> > > >
> > > >
> > > > Imagine you have an ABox A containing the following assertions:
> > > >
> > > > (1)  hor(_:1,_:2)
> > > > (2)  ver(_:2,_:3)
> > > > (3)  ver(_:1,_:4)
> > > > (4)  hor(_:4,_:5)
> > > > (5)  _:3 != _:5
> > > >
> > > > Under the traditional semantics, anonymous individuals are actually
> > > > existentially quantified variables. Hence, the ABox A is
> > > > actually equivalent to the following first-order formula \varphi:
> > >
> > > Hello,
> > >
> > > But is it possible to have the above statements in OWL 1.1? Since _:3
> > and
> > > _:5 are anonymous how can you refer to them in this difference assertion
> > > (!=)?
> > >
> > > Greetings,
> > > -gstoil
> 
> 
Received on Friday, 9 November 2007 11:15:55 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:13:27 GMT