W3C home > Mailing lists > Public > semantic-web@w3.org > May 2009

Re: OWL and LOD

From: Hugh Glaser <hg@ecs.soton.ac.uk>
Date: Wed, 13 May 2009 00:59:14 +0100
To: Toby Inkster <tai@g5n.co.uk>, Steve Harris <steve.harris@garlik.com>
CC: John Goodwin <john.goodwin@ordnancesurvey.co.uk>, "public-lod@w3.org" <public-lod@w3.org>, "semantic-web@w3.org" <semantic-web@w3.org>
Message-ID: <EMEW3|73e376a58cfcc636cb2aa51c831132e5l4C0xP02hg|ecs.soton.ac.uk|B0B1%hg@ecs.soton.ac.uk>
On 12/05/2009 11:59, "Toby Inkster" <tai@g5n.co.uk> wrote:

> On Tue, 2009-05-12 at 11:37 +0100, Steve Harris wrote:
>> I was thinking more of this issue:
>> http://lists.w3.org/Archives/Public/semantic-web/2009May/0071.html

>> re. slide 26. I've seen this done too, and it's quite concerning.
> owl:sameAs is a very blunt tool
True, and something that has been a worry for quite a while.
> and rarely useful. Instead we should be
> thinking about domain-specific equivalency predicates.
But generating equivalency predicates for every conceivable domain is not
the way to go, I would suggest.
Representing knowledge about the policy used to decide equivalence (using a
small number of equivalency predicates), context (whatever that might mean),
as well as how to interpret the equivalence in specific domains, is the
flexible and scalable approach to which we should aspire.
And if we can't begin to represent such knowledge using our knowledge
representation tools, then we should perhaps pack up and go home.
(As discussed at the LDOW2009 workshop, in fact.)
> Take for example:
>         1. <http://tobyinkster.co.uk/#i>
>         2. <http://www.doc.ic.ac.uk/~tai99/foaf.xml#me>
> Both are URLs representing me -- the first is recent and kept up to
> date; the second is a much older URL that 404s and I'm no longer able to
> update.
> Depending on what you're doing, it might be useful to treat these two as
> equivalent. If you have information on [2]'s blood type, then it's
> probably equally applicable to [1].
> But for other purposes it's useful to treat them separately. The marital
> status of [1] and [2] differ, as do their interests, their e-mail
> addresses and various other details.
> This is where equivalency predicates come in. It might be useful, for
> example, to have a foaf:samePersonAs predicate to allow us to assert
> that two URLs represent the same person, but still not glue them
> together as tightly as owl:sameAs does, so that we can still make
> statements about them individually.
> I don't know if OWL2 could then allow us to then define which of our
> favourite predicates can be cross-applied between two resources which
> are foaf:samePersonAs each other, and which may differ. (I don't think
> OWL1 is capable of that.)
> --
> Toby Inkster <mail@tobyinkster.co.uk>

Received on Wednesday, 13 May 2009 00:00:41 UTC

This archive was generated by hypermail 2.4.0 : Tuesday, 5 July 2022 08:45:12 UTC