Re: owl:sameAs [recipe]

On Tue, Jul 28, 2009 at 10:49 AM, <bill.roberts@planet.nl> wrote:
> Hi Alan
>
> Just to check I understand your point- because it's perfectly possible I am
> missing something important :-)  I'm asking for clarification here, not
> making a counter-argument.
>
> The "standard" Linked Data approach, as I understand it, is to assign HTTP
> URIs to non-information resources, ie actual things like authors, and to
> allow dereferencing of those URIs to return an information resource (eg a
> bunch of RDF triples) about the thing identified by that URI.

Yes.

> One approach
> to different data publishers having their own data on the non-information
> resource is for them to assign their own URI for it in a domain that they
> control and respond appropriately to requests to that URI.

It is important that they understand what it is that they assign their
own URI to, and there is also the principal that one should try to
reuse "vocabulary". So although I recognize that people sometimes
exercise this "choice" to publish information about things that are
already named on the web, I don't think it's a good one.

> Is your recommendation that publisher A and publisher B both publish their
> own information resources, using their own URIs of course, and to state in
> their RDF that those information resources are about (foaf:primaryTopic of)
> the relevant non-information resource (eg
> http://dbpedia.org/resource/Thomas_More from the original post) ?

Yes. There is another question about whether Dbpedia ought, in cases
where there are pre-existing URIs for entities, follow the same
practice. I think they should. But nonetheless dbpedia is fairly
unique in the world, and we ought to be trying to keep creation of
URIs that represent the same thing to a minimum. So reusing their
URI's, if you are in the LOD culture, seems like a good idea.

> In many ways this makes a lot of sense to me - but it appears to break the
> 'browseable graph'  aspect of linked data, or at least that publisher A
> loses the ability to control or contribute to what a user gets if he follows
> a link to a non-information resource URI in someone else's domain.  ie if
> the dbpedia Thomas More appears in publisher A's graph and the user (or user
> agent) dereferences it, then they get the dbpedia information about Thomas
> More, not publisher A's information about Thomas More.

This is entirely dependent on the browser technology. So far we've
been using *very* simplistic browsers on this stuff.

> Of course, that
> browseability is not the only thing to consider and publisher A's
> information about Thomas More could be found by the user through a search
> engine.

Or by smarter browsers. Remember, in the use case you are describing
you are currently looking at the manifestation and are considering
clicking on the creator link. What's to say that a browser shouldn't
notice that within the same RDF there is another triple that says:

<author> foaf:page <page about author in domain> (as my example did),

and decide that when clicking on the author link it should create a
page that shows that and *all* the foaf:pages for the author it knows
about. Or present a synthesis of them, in addition to (or instead of)
the representation returned from the request to the dbpedia-named
resource.

I sure hope browsers on the web of data are going to do a lot more for
us than they mostly do now. The current ones are useful for debugging,
but aren't terribly useful past that, for the most part, at least not
much more useful than a web page.

-Alan


>
> Cheers
>
> Bill
>
> ________________________________
> Van: Alan Ruttenberg [mailto:alanruttenberg@gmail.com]
> Verzonden: di 28-7-2009 16:02
> Aan: bill.roberts@planet.nl
> CC: public-lod@w3.org
> Onderwerp: Re: owl:sameAs [recipe]
>
> On Tue, Jul 28, 2009 at 7:58 AM, <bill.roberts@planet.nl> wrote:
>> Alan - I agree with not creating new URIs unnecessarily,
>
> It is worse than unnecessary. The statements made with them are too
> often just plain wrong. They reflect poor understanding of knowledge
> representation, of ontology, and of web architecture.
>
>> but if you are
>> setting up your own knowledge base on a topic and want to add new
>> information about those resources, and you want the user to be able to
>> dereference the resource URI to find that information, then it seems to me
>> that you pretty quickly get to the point where you have a decent reason to
>> create your own URIs.
>
> Sure, for some other kind of thing than Authors - for resources that
> are information *about* authors. If you do this you should say so. (in
> your RDF). By using foaf:primaryTopic in my rewrite of the recipe I
> said this explicitly.
>
>> This will of course lead over time to a lot of owl:sameAs links all over
>> the
>> place, but I think we just have to deal with that.
>
> We don't and I predict we won't. This practice will be a dead end. At
> least for purposes of the using the semantic web in a different way
> than the document web. I don't think we need another shadow web that
> is the same as what we have already. We need a supplement to the
> document web that lets us do new, novel things, or that lets us do
> things that are too hard to do with existing tech.
>
>> Without wanting to re-open an old thread, I broadly agree with Richard
>> Cyganiak's viewpoint in
>> http://lists.w3.org/Archives/Public/public-lod/2009Feb/0010.html
>
> And I do not, of course. That thread shows a distinct lack of vision,
> as well as a clear lack of experience in actually depending on
> knowledge representation to do anything of importance.
>
> -Alan
>
>>
>> Cheers
>> Bill
>>
>> ________________________________
>> Van: public-lod-request@w3.org namens Alan Ruttenberg
>> Verzonden: di 28-7-2009 3:47
>> Aan: Eric Lease Morgan
>> CC: public-lod@w3.org
>> Onderwerp: Re: owl:sameAs [recipe]
>>
>>
>>> --
>>
>> Sorry to be critical, but there is an important principle here: First
>> do no harm. Willy nilly inventing URIs and entities when there are
>> perfectly good ones in existence is not "cool" from a semweb point of
>> view, and unnecessary use of sameAs is both burdensome and likely to
>> lead to gross errors, as I have pointed out in my previous emails.
>>
>> -Alan
>>
>>
>

Received on Thursday, 30 July 2009 03:52:52 UTC