Re: PROV-ISSUE-370 (tracedTo-inference-only): Should tracedTo be moved to prov-constraints and be defined as a binary relation that can be inferred [prov-dm]

Hi James

I think maybe we are talking across paths.

As it is currently in the dm there is a concept called tracedto that is transitive and has attibutes. I'm just arguing for keeping it as such.

Paul

On Apr 30, 2012, at 16:07, James Cheney <jcheney@inf.ed.ac.uk> wrote:

> OK.  I interpreted Luc's request for a use case as meaning "do we have a use-case driven reason to keep it in the standard".  If it's not in the standard, then there is no reason to believe that all systems will recognize it or treat it as transitive.  
> 
> I was just pointing out that I'm not sure the appropriate constraints on tracedTo assertions are well-understood, thus working them out may represent an additional complexity burden.
> 
> --James
> 
> On Apr 30, 2012, at 2:48 PM, Paul Groth wrote:
> 
>> Err ... I wasn't suggesting that this be part of the standard. I thought Luc was asking where I would use an asserted tracedto edge and this is my use case.
>> 
>> It's nice to have an edge that I know will be treated as transitive by other systems and it's nice to be able to put attributes on it.
>> 
>> 
>> Paul
>> 
>> On Apr 30, 2012, at 13:48, James Cheney <jcheney@inf.ed.ac.uk> wrote:
>> 
>>> The potential problem with this is that such an assertion could correspond to many different actual situations.  So, if I send you a PROV instance that says
>>> 
>>> tracedTo(foo,bar,[confidence=0.5])
>>> 
>>> and nothing else, what does this mean?  Is it legal to just say tracedTo on its own?  What constraints or inferences apply?
>>> 
>>> This seems to be a way of expressing metadata/beliefs about provenance data, rather than expressing the data itself.  One could do this as an overlay on PROV without requiring everyone that uses PROV to support it, I think.
>>> 
>>> --James
>>> 
>>> On Apr 30, 2012, at 12:31 PM, Paul Groth wrote:
>>> 
>>>> My concrete use case would be to put things like confidence values on
>>>> these links. For example, in one of our systems we "guess" if there is
>>>> a tracedto and what to put some confidence value on that link. This is
>>>> one of the reasons I like attributes in the model.
>>>> 
>>>> We could do this with derivation so it's not a big deal but the nice
>>>> thing is that traced to is transitive...
>>>> 
>>>> cheers
>>>> Paul
>>>> 
>>>> 
>>>> 
>>>> On Mon, Apr 30, 2012 at 1:26 PM, Luc Moreau <L.Moreau@ecs.soton.ac.uk> wrote:
>>>>> Hi Paul,
>>>>> 
>>>>> Do you have a concrete use case, in particular, with attributes?
>>>>> 
>>>>> Thanks,
>>>>> Luc
>>>>> 
>>>>> On 04/30/2012 12:17 PM, Paul Groth wrote:
>>>>>> I think traced-to is useful to sometime assert especially in the case
>>>>>> where you want to be very vague about provenance. It's also nice to
>>>>>> have attributes so that you can associate other sorts of information
>>>>>> with it.
>>>>>> 
>>>>>> However, if others think it's nicer to be inference only then I won't
>>>>>> be stand in the way.
>>>>>> 
>>>>>> cheers
>>>>>> Paul
>>>>>> 
>>>>>> On Mon, Apr 30, 2012 at 1:02 PM, Provenance Working Group Issue
>>>>>> Tracker<sysbot+tracker@w3.org> wrote:
>>>>>> 
>>>>>>> PROV-ISSUE-370 (tracedTo-inference-only): Should tracedTo be moved to prov-constraints and be defined as a binary relation that can be inferred [prov-dm]
>>>>>>> 
>>>>>>> http://www.w3.org/2011/prov/track/issues/370
>>>>>>> 
>>>>>>> Raised by: Luc Moreau
>>>>>>> On product: prov-dm
>>>>>>> 
>>>>>>> 
>>>>>>> TracedTo was introduced in the data model so as to have a transitive relation over derivations, etc. It can be inferred. In contrast, its definition as an assertion was not very compelling. In the latest version of prov-constraints, it is only defined as something that can be inferred.
>>>>>>> 
>>>>>>> Really, it looks like a relation that is useful to express queries.
>>>>>>> 
>>>>>>> So, in the spirit of simplification, should we move it out of prov-dm, and have it defined in prov-constraints only.
>>>>>>> 
>>>>>>> At the same time, it could be simplified to a binary relation, since we have no way of inferring attributes for this relation.
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>> 
>>>>>> 
>>>>>> 
>>>>> 
>>>>> --
>>>>> Professor Luc Moreau
>>>>> Electronics and Computer Science   tel:   +44 23 8059 4487
>>>>> University of Southampton          fax:   +44 23 8059 2865
>>>>> Southampton SO17 1BJ               email: l.moreau@ecs.soton.ac.uk
>>>>> United Kingdom                     http://www.ecs.soton.ac.uk/~lavm
>>>>> 
>>>>> 
>>>> 
>>>> 
>>>> 
>>>> -- 
>>>> --
>>>> Dr. Paul Groth (p.t.groth@vu.nl)
>>>> http://www.few.vu.nl/~pgroth/
>>>> Assistant Professor
>>>> Knowledge Representation & Reasoning Group
>>>> Artificial Intelligence Section
>>>> Department of Computer Science
>>>> VU University Amsterdam
>>>> 
>>>> 
>>> 
>>> 
>>> -- 
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>> 
>> 
>> 
> 
> 
> -- 
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
> 

Received on Monday, 30 April 2012 14:42:01 UTC