Afraz Jaffri wrote:
We are pleased to announce a Linked Data site for the Ordnance Survey,
available at:
http://os.rkbexplorer.com
with links from over 8000 URIs to Geonames URIs. Take 'Hampshire' as
an example:
Afraz,
http://os.rkbexplorer.com/description/osr7000000000017765
kidehen$ curl -I -H "Accept: application/rdf+xml"
http://os.rkbexplorer.com/description/osr7000000000017765
HTTP/1.1 406 Not Acceptable
Date: Sat, 12 Jul 2008 20:31:35 GMT
Server: Apache/2.0.52 (Red Hat)
X-Powered-By: PHP/4.3.9
Content-Type: text/html; charset=UTF-8
Connection: close
Why no 303 or use of <link rel="[relavant-predicate]" .../> so that
User Agents can locate
<http://os.rkbexplorer.com/data/osr7000000000017765> .
My issues are summarized here:
http://demo.openlinksw.com/rdfbrowser2/?uri%5B%5D=http%3A%2F%2Fos.rkbexplorer.com%2Fdescription%2Fosr7000000000017765&uri%5B%5D=http%3A%2F%2Fos.rkbexplorer.com%2Fdata%2Fosr7000000000017765&;
Kingsley
It might also be the right time, with all the owl:sameAs discussion,
to practically demonstrate how our coreference system works. In the
above example the link created is between 'Hampshire the county' from
the OS and 'Hampshire the second order administrative division' from
Geonames. We do not know if these two entities are exactly the same,
so
instead of using owl:sameAs we use our own coref:duplicate predicate.
One of the features of our system is that knowledge about coreference
is separated from the knowledge of the actual entity. In the RDF for
the above URI at http://os.rkbexplorer.com/data/osr7000000000017765
you
will find:
<coref:coreferenceData rdf:resource="http://os.rkbexplorer.
com/crs/osr7000000000017765"/>
Resolving this URI will give you a 'bundle' containing the duplicates:
<coref:Bundle>
<coref:canon rdf:resource="http://os.rkbexplorer.
com/id/osr7000000000017765"/>
<coref:duplicate rdf:resource="http://os.rkbexplorer.
com/id/osr7000000000017765" />
<coref:duplicate rdf:resource="http://sws.geonames.org/2647554/"/>
<coref:lastUpdated>2008-07-10 11:39:44</coref:lastUpdated>
</coref:Bundle>
As you can see one URI is chosen as the canonical URI to use. The
separation of coreference means, to a limited extent, that the
context
of duplication can be preserved. If I wanted to say that under some
other context there were other URIs that were deemed to be the same
then I can simply create another bundle with another <coref:
coreferenceData> predicate in the RDF for the entity. Of course, the
question of how to show the context is yet to be solved...
There may be some errors in the equivalences. All feedback is
greatfully received :)
In relation to another question about how owl:sameAs is currently
being used, there are some examples in our LDOW paper:
http://eprints.ecs.soton.ac.uk/15181/
in particular http://dbpedia.org/resource/Welsh and http://dbpedia.
org/resource/Lilac
Regards,
Afraz
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On
Behalf Of Harry Halpin
Sent: 09 July 2008 10:55
To: Hugh Glaser
Cc: Bijan Parsia; Peter Ansell; semantic-web at W3C; [EMAIL PROTECTED]
org
Subject: Re: How do you deprecate URIs? Re: OWL-DL and linked data
Hugh Glaser wrote:
Thanks guys, a really interesting and important discussion.
However, after the last couple of postings I have the feeling I
may
agree
with both of you.
Is that possible?
Bijan et. al. are right about the semantics of owl:sameAs, but as
I've
said before, I think that something weaker needs to be coined
("lod:equivalentTo") that states that two URIs refer to the same
thing
but that any semantic entailments *may* not hold (i.e. user beware).
That's a dangerous thing, I agree, but it seems to be what the
Linked
Data community needs and what's happening organically in the wild
with
the (ab)use of owl:sameAs.
Hugh
__________________________________________________________
Free games for a wet weekend - http://www.tiscali.co.uk/play
__________________________________________________________
__________________________________________________________
Free games for a wet weekend - http://www.tiscali.co.uk/play
__________________________________________________________
--
Regards,
Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com
