Re: prov-o review / comments

On Mar 6, 2012, at 1:10 PM, Paul Groth wrote:

> Hi Khalid 
> 
> I think the inference in the dm is about the properties - not their reifications per say

Agreed.



> 
> Paul
> 
> On Mar 6, 2012, at 18:31, Khalid Belhajjame <Khalid.Belhajjame@cs.man.ac.uk> wrote:
> 
>> 
>> Hi Tim,
>> 
>> I agree that the properties can be free to differ from the "Involvement" 
>> hierarchy.
>> 
>> Regarding the flattening that you are suggesting in the Involvement 
>> hierarchy, I am wondering if it may yield some issues later on. In 
>> particular, if there are people who want to inject some inference rules 
>> (constraint) in the ontology.
>> For example, an inference rule that can be 
>> applied to prov:Association should be also applicable to prov:End and 
>> prov:Start (according to the DM), but the flattening suggested will 
>> remove that implication.

Since we are being _less_ specific in the proposed hierarchy, they are free to provide the [:End subclassOf :Association] if they need it.
I am betting that most will not need :End to be an :Association, but what they really need to know is that it qualifies the involvement of an Agent (AgentInvolvement).


>> I don't think that the issue I am raising is 
>> blocking, but I would like to know if people already thought of it.

The model in DM is upheld by the subproperties, and the Involvements still provide the trigger to the binary relations.
These are the two most important aspects w.r.t. modeling.
Meanwhile, I'm trying to gain some simplicity and readability at the (comparatively small) expense of requiring a reasoning expert to include a subclass relation, which they'll know how to handle.

Regards,
Tim



>> 
>> Thanks, khalid
>> 
>> 
>> On 06/03/2012 16:15, Timothy Lebo wrote:
>>> prov-wg,
>>> 
>>>> However the newer, split DM has changed some of these semantics, I am
>>>> not now (quickly) able to find any relation subtypes that cause
>>>> 'inheritence' of attributes and record id. The DM constraints [2] does
>>>> not seem to inherit attributes, but allow 'any' attributes ("for some
>>>> gAttr") in the inferred relations, except for this - perhaps strange
>>>> one:
>>>> 
>>>> If the records entity(e,attrs) and wasAssociatedWith(a,e) hold for
>>>> some identifiers a, e, and attribute-values attrs, then the record
>>>> agent(e,attrs) also holds. So to be WD4 compliant we should not have
>>>> any hierarchy of prov:Involvement beyond them being involvements.
>>> 
>>> For the sake of simplicity, I would like to propose that we follow Stian's suggestion regarding the subclass hierarchy under Involvement.
>>> The critical aspect that we are conveying with the Involvement hierarchy is that we are referencing some binary relation to an Activity, Entity, or Agent.
>>> Anything further is not provided by the hierarchy, at the cost of confusion.
>>> 
>>> Does anyone have an objection to flattening the hierarchy to "stop" at the primary Elements (Activity, Entity, Agent)?
>>> 
>>> prov:Involvement
>>>    prov:ActivityInvolvement
>>>        prov:Generation
>>>        prov:Inform
>>>        prov:StartByActivity
>>>    prov:EntityInvolvement
>>>        prov:AgentInvolvement
>>>            prov:Association
>>>            prov:End               # This raised a level
>>>            prov:Start              # This raised a level
>>>            prov:Attribution
>>>            prov:Responsibility
>>>        prov:Derivation
>>>        prov:Source         # This raised a level
>>>        prov:Revision      # This raised 2 levels
>>>        prov:Quotation
>>>        prov:Usage
>>>    prov:Trace  # This raised a level (b/c it refers to either Activities or Entities)
>>> 
>>> The property hierarchy would be free to differ from the class hierarchy.
>>> 
>>> In the absence of objections, I will make the change by the end of the week.
>>> 
>>> Regards,
>>> Tim
>>> 
>>> 
>>> 
>>>> Luc - is this the correct interpretation?
>>>> 
>>>> [1] http://www.w3.org/TR/2012/WD-prov-dm-20120202/
>>>> [2] http://dvcs.w3.org/hg/prov/raw-file/default/model/prov-dm-constraints.html
>>>> 
>>>> 
>>>> -- 
>>>> Stian Soiland-Reyes, myGrid team
>>>> School of Computer Science
>>>> The University of Manchester
>>>> 
>>>> 
>>> 
>>> 
>> 
>> 
>

Received on Tuesday, 6 March 2012 18:31:30 UTC