Re: Dataset vocabularies vs. interchange vocabularies

Kevin Richards wrote:
> This mention of owl:sameAs reminds me of the mention of the "sameAs issue", at ISWC, that has developed in the semantic web arena.
> I can imagine what this issue is, but am not 100% sure, so can anyone explain this issue to me?
>   
I remember there is a discussion on one of the W3C's mailing list about 
the ambiguity of "owl:sameAs" but couldn't remember where it is.  But I 
can offer my own explanation.

Its cause is a very fundamental one because the current URI 
specification is, in fact, syntactically incomplete.  With current URI 
spec, the referential realm of URI is anything but URI.  (Of course, I 
can mint another URI to denote another instance of a URI class, but then 
there needs to be a standard way to find this information). 

Thus, if a URI denotes a URI, the semantics of "a owl:sameAs b" would 
mean that "a" is simply an syntactic alias of "b" and nothing-else.   
But if a URI denotes a non-URI resource, "a owl:sameAs b" would imply 
that a semantic alias.  That is: "a" is an syntactic alias of "b" PLUS 
a's and b's representations collectively describe the meaning of "a/b".  
To put it plainly, the second semantics implies an owl:import but the 
first one does not because the representation of a URI is literally 
itself but the representation of a resource requires dereference.  I 
think, with the current URI specification, the owl:sameAs is for the 
second semantics.  In this case, it is not as cheap as people think that 
minting a URI first and later binding it with others using owl:sameAs. 
If a's representation is not logically consistent with b's, then the 
owl:sameAs leads to a null model.

What makes the issue complicated, however, is that the ambiguity of 
owl:sameAs is not with the problem of OWL but that of URI.  The issue 
also touched upon our conceptualization of URI, resource, 
representation, information/document, and meaning.  I have discussed 
this issue quite extensively in a manuscript that I have submitted to 
WWW2009.  As a pending manuscript, I don't think that it is appropriate 
to post it here.  But if anyone is interest in reading it, send me an email.

Xiaoshu




> Thanks
> Kevin Richards
>
> -----Original Message-----
> From: semantic-web-request@w3.org [mailto:semantic-web-request@w3.org] On Behalf Of Harry Halpin
> Sent: Friday, 28 November 2008 2:05 a.m.
> To: Richard Cyganiak
> Cc: John Graybeal; public-lod@w3.org; Semantic Web
> Subject: Re: Dataset vocabularies vs. interchange vocabularies
>
>
> On Thu, 27 Nov 2008, Richard Cyganiak wrote:
>
>   
>> On 26 Nov 2008, at 21:53, John Graybeal wrote:
>> <snip>
>>     
>>> would you agree that duplicating a massive set of URIs for 'local technical
>>> simplification' is a bad practice? (In which case, is the question just a
>>> matter of scale?)
>>>       
>> You are asking me if 'local technical simplification' is a good reason or a
>> bad reason for duplicating URIs? Uh, I guess it depends...
>>
>> My point was this: The key benefits of URI re-use can also be obtained by
>> minting your own URIs and linking them to existing URIs via adequate RDF
>> properties. And that practice can have additional practical/implementation
>> benefits (and costs). Hence, consider both options; there's no reason to
>> knee-jerk against creating new identifiers.
>>     
>
> I agree in theory with Richard, but in practice with John. The key
> benefits of URI re-use can  only be gained by using multiple URIs if we
> have "adequate URI properties"  (i.e. owl:sameAs?) and given  an adequate
> reasoning system  that can identify the same URIs in any data  set -
> including large ones -  where we want to merge data using these  "inferred
> to be the same" URIs.
>
> To my knowldge, we have neither adequate URI properties or working
> reasoning services, at least for the end-user. Now perhaps this will
> change, but if not, why not re-use URIs?
>
> If we do have adequate URI properties besides the infamous owl:sameAs,
> please point me to them. And while at ISWC there was clearly lots of work
> on large-scale identity management trying to discover URI equivalences via
> inference, I'm not sure how well that works right now.
>
> Furthermore, there's the question of what URI  to use in the output if one
> is identifying URI's to be the same and one wants to re-use the merged data.
>
>                 -harry
>
>
>   > Best,
>   
>> Richard
>>
>>
>>
>>     
>>> John
>>>
>>> --------------
>>> John Graybeal   <mailto:graybeal@mbari.org>  -- 831-775-1956
>>> Monterey Bay Aquarium Research Institute
>>> Marine Metadata Interoperability Project: http://marinemetadata.org
>>>       
>
> --
>                                 --harry
>
>         Harry Halpin
>         Informatics, University of Edinburgh
>          http://www.ibiblio.org/hhalpin
>
>
> Please consider the environment before printing this email
> Warning:  This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
> The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz
>
>   

Received on Monday, 1 December 2008 10:30:02 UTC