Re: PROV-ISSUE-222 (used-objectproperty): Datatype property for used? [Ontology]

Hi All 

Since I raised the issue, let me get an example up ... I think it will be a no brainer to solve just really need a good place to put application specific descriptions for example command line arguments.

Example coming tomorrow

Paul


On Apr 16, 2012, at 19:45, Timothy Lebo <lebot@rpi.edu> wrote:

> 
> On Apr 16, 2012, at 1:31 PM, Stephan Zednik wrote:
> 
>> 
>> On Apr 16, 2012, at 11:15 AM, Timothy Lebo wrote:
>> 
>>> 
>>> On Apr 16, 2012, at 11:44 AM, Stian Soiland-Reyes wrote:
>>> 
>>>> prov:value can specialize rdf:value ( and standards say so), but for is it would not really add any meaning beyond anything given by its domain (say prov:Entity).
>>>> 
>>> 
>>> I don't see the need to mirror it when rdf:value works just fine and already recognized by so many tools.
>> 
>> While rdf:value is recognized by tools, it has no defined meaning on its own (according to http://www.w3.org/TR/rdf-schema/#ch_value).
> 
> Thanks for pointing this out.
> "rdf:value is an instance of rdf:Property that may be used in describing structured values."
> is NOT how rdf:value has come to be used in the wild.
> 
> Funny that the example that they cite doesn't use rdf:value …. http://www.w3.org/TR/2004/REC-rdf-primer-20040210/#example16
> 
> So then I'd propose we make prov:value a DatatypeProperty and provide a better definition than what the RDF spec provided.
> 
> 
>>  I also believe direct usage without restricting its type to owl:ObjectProperty or owl:DatatypeProperty also puts an ontology into OWL Full.
> 
> Another huge reason to define our own :-)
> 
>> 
>>> 
>>>> But we want string activities as well?
>>>> 
>>>> 
>>> 
>>> That's impossible. (and one says that, it means they should make an axiom…. prov:value rdfs:domain prov:Entity (which is disjoint with Activity))
>>> But worth it's weight of another property?
>> 
>> It seems to me we are conflating simple descriptions of activities and entities with the actual activity and entity resource.
> 
> I agree that this conflation is bad, but I can't say that I'm seeing it.
> Perhaps it's because there isn't an example on this issue yet.
> 
>> 
>> Why not just have an annotation that provides a human-readable description of the activity or entity?
>> 
>> To replace Activity/Entity individuals with string descriptions of said individuals would be a mistake.
> 
> +100
> 
> -Tim
> 
> 
>> 
>> --Stephan
>> 
>>> 
>>> 
>>>> We should be careful not to overlap rdfs:label...
>>>> 
>>>> 
>>> 
>>> Who proposed using rdfs:label?
>>> Agreed, this should be left out of the discussion.
>>> 
>>> -Tim
>>> 
>>> 
>>> 
>>> 
>>>> On Apr 16, 2012 4:36 PM, "Timothy Lebo" <lebot@rpi.edu> wrote:
>>>> 
>>>> On Apr 16, 2012, at 10:49 AM, Luc Moreau wrote:
>>>> 
>>>> > Hi Tim,
>>>> >
>>>> > Just a word to say that it's a problem that is not specific to the ontology.
>>>> > The problem is similar in other serializations.
>>>> > Should we have a statement about this in the dm?
>>>> 
>>>> That makes sense. Would you life to reserve prov:value?
>>>> PROV-O will not define prov:value in favor of rdf:value.
>>>> I think the rest of the PROV-O solution (content in RDF vocab) would fall outside of DM's control, as we've done before.
>>>> 
>>>> -Tim
>>>> 
>>>> > Luc
>>>> >
>>>> > On 04/16/2012 02:18 PM, Timothy Lebo wrote:
>>>> >> Paul (and Graham),
>>>> >>
>>>> >> The prov-o team discussed this last week and agreed that this topic is more appropriate in the best practices document.
>>>> >> We also outlined the recommended patterns.
>>>> >>
>>>> >> I put a stub entry at
>>>> >>
>>>> >> http://dvcs.w3.org/hg/prov/raw-file/1a7d883e143e/bestpractices/BestPractices.html#using-strings
>>>> >>
>>>> >> that says:
>>>> >>
>>>> >> * If you want to break RL and any tools built around PROV-O, just use a string.
>>>> >> * If you want to follow the datatype/objectproperty distinction, use a resource with rdf:value OR
>>>> >> * use content in rdf http://www.w3.org/TR/Content-in-RDF10/
>>>> >>
>>>> >> 1)
>>>> >> Can we move this issue to the best practices product?
>>>> >> http://www.w3.org/2011/prov/track/products/7
>>>> >>
>>>> >> 2)
>>>> >> Can you put a "string-heavy" example into http://www.w3.org/2011/prov/wiki/PROV_examples to motivate further development of the best practice?
>>>> >>
>>>> >> 3)
>>>> >> Can we close ISSUE-248 http://www.w3.org/2011/prov/track/issues/248 as a duplicate of this issue?
>>>> >>
>>>> >>
>>>> >> On Jan 19, 2012, at 4:36 AM, Graham Klyne wrote:
>>>> >>
>>>> >>
>>>> >>> Paul,
>>>> >>>
>>>> >>> This problem is, IMO, an atifact of the arguably arbitrary restrictions of description logic and OWL-DL.  If you don't need to be consrainted to OWL-DL then the problem does not arise.  Just saying.
>>>> >>>
>>>> >> The problem does arise practically, too. If the range of prov:used is a rdfs:Resource, then tools will handle it as such (and not a string).
>>>> >> So tools will choke while reading your account, even if they don't care about reasoning.
>>>> >>
>>>> >>
>>>> >>
>>>> >>> Staying with the object/datatype property distinction, I think either of your suggested approaches can work, but I don't know about semantics of entity here - it seems to me that it should be possoible to formulate the semantics around two properties as well as one, even if the formulation is more complex.
>>>> >>>
>>>> >>
>>>> >>
>>>> >>> The second approach avoids the semantic uncertainties at the costof some added complexity in RDF representation.
>>>> >>>
>>>> >>
>>>> >> @Graham, could you elaborate this approach, so that we can articulate it in the best practices document?
>>>> >>
>>>> >> Thanks,
>>>> >> Tim
>>>> >>
>>>> >>
>>>> >>
>>>> >>
>>>> >>> I'm not sure this helps :(
>>>> >>>
>>>> >>> #g
>>>> >>> --
>>>> >>>
>>>> >>> On 18/01/2012 09:40, Provenance Working Group Issue Tracker wrote:
>>>> >>>
>>>> >>>> PROV-ISSUE-222 (used-objectproperty): Datatype property for used? [Ontology]
>>>> >>>>
>>>> >>>> http://www.w3.org/2011/prov/track/issues/222
>>>> >>>>
>>>> >>>> Raised by: Paul Groth
>>>> >>>> On product: Ontology
>>>> >>>>
>>>> >>>> Currently, prov-o:used is defined as an objectproperty. This is fine. However, we've be doing some modeling here at the VU where the parameter to a program is a string. Currently, this is not modelled using a prov-o:used edge but it seems like it should be. Is there anyway we can support this?
>>>> >>>>
>>>> >>>> My first inclination is to define a corresponding datatype property but this make break the semantics of entity...
>>>> >>>>
>>>> >>>> Another option might be to suggest using a blank node with the string attached using an application specific predicate.
>>>> >>>>
>>>> >>>> Suggestions?
>>>> >>>>
>>>> >>>>
>>>> >>>>
>>>> >>>>
>>>> >>>>
>>>> >>>
>>>> >>>
>>>> >>
>>>> >>
>>>> >
>>>> > --
>>>> > 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
>>>> >
>>>> >
>>>> >
>>>> 
>>>> 
>>> 
>> 
>