OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

xdi message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: Re: [xdi] reactions on paper "When owl:sameAs isn?t the Same..."


Dear Paul, Drummond, et al.

in thanking you for your many replies in this productive discussion,  
I'd like to say that my original intention was to discuss how the TC  
could address at least some of the four (mis)use cases addressed in  
the original paper (http://www.w3.org/2009/12/rdf-ws/papers/ws21),  
with particular focus on case #2. The four (mis)use cases were:

1. Same Thing As But Referentially Opaque - two URIs do refer to the  
same thing, but all the properties ascribed to one URI ARE NOT  
NECESSARILY ACCEPTED BY THE OTHER.

I think this is the use case you're currently discussing, Drummond's words:

>> "an identity relation between two XDI subjects in which there is  
>> not a requirement that all properties of one subject apply to the  
>> other and vice versa, nor a requirement that for XDI statement  
>> consistency across their respective XDI graphs."

h:correlation as you describe Paul I think applies here - paper  
authors also suggest skos:closeMatch

2. Same Thing As But Different Context - This is a DIFFERENT use case  
were two URIs do refer to the same thing AND ALL PROPERTIES DO HOLD  
FOR BOTH, BUT WE CANNOT REUSE THE URI IN A DIFFERENT CONTEXT.

For this use case, authors suggest to use named graph. My original  
point is all here: they have to use named graph as URIs are opaque by  
W3C definition, but XRIs are not. So to me, XDI nicely addresses this  
case, as we have a way to explicitly state context in XRIs:

=Lynn.Stein
=Rachel.Stein+mom=Lynn.Stein <!-- at Parent Teacher Association
@MIT+professor=Lynn.Stein <!-- at MIT

note that

=Rachel.Stein+mom=Lynn.Stein/$is$a/=Lynn.Stein

and

@MIT+professor=Lynn.Stein/$is$a/=Lynn.Stein

as from definition of $has and composite identifiers already agreed in the TC.

NOTE: privacy issue: during last phc Bill correctly stated that using  
@MIT+professor=Lynn.Stein or @golfclub+member=Drummond could be a  
privacy violation in some cases. Therefore !number has been proposed  
instead. My point is that the relation between @golfclub+member!43 and  
=Drummond is the following

@golfclub+member!43/$is/@golfclub+member=Drummond
@golfclub+member=Drummond/$is$a/=Drummond

Obviously these two statements - especially the first one - should be  
protected under link contracts in privacy aware applications.
Now a shortcut for this COULD be =Drummond/$as/@golfclub+member!43,  
but it is not absolutely needed.

Other two (mis)use cases, less interesting for us in my opinion:

3. Represents - when identity is conflated with representation. I  
think a proper use of owl:inverseFunctionalProperty would solve the  
problem here.

4. Very Similar To - mostly a subcase of 1.

Thank you and, as usual, your comments are very welcome!

Kind Regards,
Giovanni

Def. Quota "Paul Trevithick" <ptrevithick@gmail.com>:

> Thanks Drummond.
>
> Here are a few more thoughts:
>
> 1) I think that h:correlation (c. 2005) is the same as  
> oguid:identical [1] (c. 2008) although I've not studied that too  
> carefully. A quote from [1]:
>
> The property oguid:identical indicates two resources are  
> co-referent. In other words, they refer to the same real world thing.
>
> Open GUID identity is defined pragmatically as referring to the same  
> concept beyond a reasonable doubt. The major determinant is if the  
> average human perceives the resources as referring to the same thing.
>
> The linking implication can be described as being between OWL sameAs  
> and RDFS seeAlso. The sameAs property declares equivalence of  
> semantic representations; thus it should be used carefully to avoid  
> confusing logical reasoners. The seeAlso property provides  
> additional information; nothing useful can be inferred from the  
> link. The identical property declares equivalence of referents; thus  
> it allows reasoners to use local representations while still  
> maintaining global context.
>
> 2) The concept of coreference [2] may also be helpful.
>
> 3) In Higgins the to co-referents must exist in different contexts.
>
> With that as background, I'd like to say that it would be great if  
> we could arrive at a precise semantic (irrespective of predicate  
> name, $ word, etc.) that we could all use. If this TC comes up with  
> a compatible but more precisely worded definition, I'd like nothing  
> more than to copy/paste that definition into [3].
>
> [1] http://openguid.net/specification#identical
> [2] http://en.wikipedia.org/wiki/Coreference
> [3] http://wiki.eclipse.org/Higgins_Data_Model_2.0#Attributes
>
> On Jun 25, 2010, at 1:03 AM, Drummond Reed wrote:
>
>> > Paul wrote:
>> > Agreed (although I think you meant "symmetric" vs. transitive)
>>
>> Paul, you're right, we did mean symmetric.
>>
>> BTW, though we spent the last half of today's telecon discussing  
>> this issue, and agreed that the full $is semantics --- i.e. the  
>> semantics of a full identity relation -- is too strong for just  
>> "observed correlation", we didn't yet get into either:
>>
>> 1) What the precise definition would be for this weaker form of  
>> identity relation.
>>
>> 2) What the XDI $word for it would be.
>>
>> I think the TC is open to suggestions for both. My suspicion is  
>> that the definition is probably along the lines of one of the four  
>> options discussed in http://www.w3.org/2009/12/rdf-ws/papers/ws21  
>> (probably "4.1 Same Thing As But Referentially Opaque" or "4.2 Same  
>> Thing As But Different Context" or some combination of these two).
>>
>> My only suggestion to a $word for this weaker form is $as. Example:
>>
>> =drummond/$as/@golfclub+member!43
>>
>> The semantics of $as would be "an identity relation between two XDI  
>> subjects in which there is not a requirement that all properties of  
>> one subject apply to the other and vice versa, nor a requirement  
>> that for XDI statement consistency across their respective XDI  
>> graphs."
>>
>> That's just a strawman - hack away.
>>
>> =Drummond
>>
>> On Thu, Jun 24, 2010 at 2:10 PM, Paul Trevithick  
>> <ptrevithick@gmail.com> wrote:
>> Agreed (although I think you meant "symmetric" vs. transitive)
>>
>> On Jun 24, 2010, at 3:22 PM, Barnhill, William [USA] wrote:
>>
>>> From my standpoint I've always thought of the XDI synonym dollar  
>>> word, which is now $is, as mapping to owl:sameAs. I briefly  
>>> thought it  might be better to map to something a little weaker,  
>>> skos:closeMatch, but came back around to believing it should map  
>>> to owl:sameAs. Another reason I think h:correlation couldn't map  
>>> to $is is that h:correlation is specifically not a transitive  
>>> property according to the Higgins XDI Harmonization wiki, whereas  
>>> $is must be.
>>>
>>> Bill
>>>
>>> From: Paul Trevithick [mailto:ptrevithick@gmail.com]
>>> Sent: Thursday, June 24, 2010 2:15 PM
>>> To: Giovanni Bartolomeo
>>> Cc: OASIS - XDI TC
>>> Subject: Re: [xdi] reactions on paper "When owl:sameAs isn?t the Same..."
>>>
>>> Giovanni,
>>>
>>> Last week it was proposed that h:correlation is the same as $is,  
>>> but on further reflection I don't think that's right. I think $is  
>>> is equivalent in semantics to owl:sameAs. Joseph and I discussed  
>>> this a bit today on the XDI harmonization portion of the weekly  
>>> Higgins call. In Higgins h:correlation means: representing the  
>>> same thing in different contexts. The following para is copied  
>>> from [1] (BTW, we use the term entity instead of resource):
>>>
>>> h:correlation is subtly different from owl:sameAs. It is statement  
>>> made by a human observer that the source and target of this link  
>>> are believed to be alternative representations of the same real  
>>> world person or object. A single, natural person would thus be  
>>> represented by different entities in different contexts. This  
>>> linkage does not presume that the entire set of attributes across  
>>> these entities, if they were brought together and combined, is  
>>> necessarily logically consistent. The ontologies in the two  
>>> contexts may be such that each of the two representations cannot  
>>> be merged and remain logically consistent. For this reason Higgins  
>>> does not use owl:sameAs which does imply this ability to directly  
>>> merge representations. h:correlation is stronger than rdfs:seeAlso  
>>> but weaker than owl:sameAs.
>>>
>>> [1] http://wiki.eclipse.org/Higgins_Data_Model_2.0
>>>
>>> On Jun 24, 2010, at 1:40 PM, Giovanni Bartolomeo wrote:
>>>
>>>> Hello,
>>>>
>>>> having had time this week to look at these Drummond suggested readings:
>>>>
>>>> ?When owl:sameAs isn?t the Same: An Analysis of Identity Links on the
>>>> Semantic Web?, by Harry Halpin, Ivan Herman, and Patrick J. Hayes
>>>>
>>>> ?RDF and XML: Towards a Unified Query Layer?, by Nuno Lopes, Stefan
>>>> Bischof, Orri Erling, Axel Polleres, Alexandre Passant, Diego
>>>> Berrueta, Antonio Campos, Jé?rôme Euzenat, Kingsley Idehen, Stefan
>>>> Decker, Sté?phane Corlosquet, Jacek Kopecky ?, Janne Saarela, Thomas
>>>> Krennwallner, Davide Palmisano, and Michal Zaremba
>>>>
>>>> (both will be presented at nextcoming W3C RDF workshop,
>>>> http://www.w3.org/2009/12/rdf-ws/) I would like to share with you some
>>>> thoughts on how I believe XDI and XRI non-opaque identifiers could
>>>> nicely address some issues presented there - especially in the first
>>>> article.
>>>>
>>>> Could you insert this topic into today's or next week's phc agenda?
>>>>
>>>> Thank you very much,
>>>> Giovanni
>>>
>>
>>
>
>



----------------------------------------------------------------
Invito da parte dell'Ateneo:
Il tuo futuro e quello della Ricerca Scientifica hanno bisogno del
tuo aiuto. Dona il  5 x mille all'Universita' di Roma Tor Vergata
codice fiscale: 80213750583 http://5x1000.uniroma2.it



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]