Re: PROV-CONSTRAINTS - how can we know they're satisfied?

On 06/06/2012 06:21, Luc Moreau wrote:
> Hi Graham,
>
> I thought that constraints can always be applied to decide whether provenance descriptions are valid or not.

If all you have is the provenance, and if you are operating in an open-world 
context (i.e. not assuming you have complete information), you can't know for 
sure that all the constraints are satisfied by the denotations of the terms used 
merely by looking some provenance.  Sure, you might be able to detect some 
violations, but you can't be sure you've got them all.

E.g. (considering Paul's original blog example) if you've used a URI for a 
resource whose content and authorship changes over time, and used provenance to 
express it's authorship.  Until someone else adds a different claim of 
authorship, you don't know that the intended invariability of the resource is 
not satisfied.

So, while you may be able to use the constraints to reason within a closed 
context, more is needed if you want to lift [1] the reasoning to a broader context.

To be clear, this is not something I think should be explicitly covered in the 
specs we're currently working on, but may be something to consider for future 
refinements.

#g
--

[1] "lift" used here in the sense described by: R. V. Guha, Contexts: A 
Formalization and Some Applications, Ph.D. Thesis, Stanford University, 1995, 
linked from http://www-formal.stanford.edu/guha/.



>
> Professor Luc Moreau
> Electronics and Computer Science
> University of Southampton
> Southampton SO17 1BJ
> United Kingdom
>
> On 5 Jun 2012, at 22:38, "Graham Klyne"<graham.klyne@zoo.ox.ac.uk>  wrote:
>
>> Something that's been gnawing at the back of my mind is a detail of how the PROV-CONSTRAINTS story plays out.  My understanding is that the data model and ontology terms are defined separately from the constraints that underpin some inferences about provenance that are believed to be useful.  But we don't want to prohibit someone from using and publishing provenance data without taking care of all the niceties (the "scruffy" approach).
>>
>> So how can we know if some arbitrary set of provenance information is intended (or claimed) to conform to PROV-CONSTRAINTS?
>>
>> It seems to me that this proposal by Pat Hayes might, if adopted by the RDF group, supply a mechanism that can be used to assert that all the terms of the provenance ontology are being used in accordance with the PROV-CONSTRAINTS, and hence can signal validity of the various inferences described in PROV-CONSTRAINTS.
>>
>>   http://www.w3.org/2011/rdf-wg/wiki/AnotherSpin
>>
>> #g
>> --
>>
>

Received on Wednesday, 6 June 2012 06:47:03 UTC