Re: Contextualization ---> Optional bundle in Specialization

On 28/06/2012 12:56, Timothy Lebo wrote:
>
> On Jun 27, 2012, at 6:31 PM, Graham Klyne wrote:
>
>> On 27/06/2012 19:21, Paul Groth wrote:
>>> So the use case is the issue?
>>>
>>> I really don't get how the example breaks any semantics. Sorry...
>>>
>>> So I think that your approach to allowing a qualified specialization would be fine with me especially if we add a inBundle predicate that identifies a bundle. but Tim was really really against this because of the increased number of triples.
>>
>> Well, my proposal tries to avoid adding a *specific* "in bundle" qualification at this time,
>
> (but that's what we're trying to say: "as I saw it in that bundle")
>
>> just a hookm for introducing one when we know how the semantics could work.
>
> There is no semantics :-)
>
>>
>> I think the "increased number of triples" is symptomatic of the problem here. The situation might (only "might" - I'm not sure) be addressed by reifying the inclusion of statements about an entity in a bundle
>
>
> ^^ This is the kind of "context" that isn't being included in the current proposal, because we're not trying to establish or preserve any "other RDF hanging around our subject". This sort of thing is exactly what we should avoid because of all of the rabbit holes that the other WGs have faced and are facing. We aren't the group to solve it.
>
>
>> ... and, as we have seen in the past, and more recently with the qualified relation patterns in PROV-O, reifying stuff within the RDF model does tend to cause an explosion of triples.
>>
>> To reiterate:  I think the appropriate and long-term solution here is to have proper semantics for RDF datasets, and then to build the contextualization of provenance around those semantics.  So far, the RDF group hasn't delivered any dataset semantics, which is unfortunate.
>
>
> I agree, if we were trying to do anything like what you describe. But we're not. We're just providing a type of specialization that specializes another entity that we've seen in a particular bundle. Three, distinct entities that are described with two triples using our own predicates.

My responses have been based on Luc's rejection of an earlier proposal of mine, 
which was very similar to what you seem to be after...  Maybe different 
proponents of the contextualizationOf proposal have different goals?

> I guess this topic has been beaten out of DM.
>
> So I'll press on in my applications with dcterms:isReferencedBy [ a prov:Bundle ] and not lose any sleep.

I think that will be fine, because it clearly doesn't try to conextualize the 
semantics associated with the occurrence.

#g
--

Received on Thursday, 28 June 2012 15:06:39 UTC