Re: DQV: Proposal to solve ISSUE-186

Hi,

On Sun, Sep 6, 2015 at 11:07 PM, Antoine Isaac <aisaac@few.vu.nl> wrote:
>
> In side discussion, Riccardo and I have identified a proposal for closing
> another issue on DQV
> https://www.w3.org/2013/dwbp/track/issues/186 - "There might be no need
> for a subclass link between dqv:QualityMeasure and daq:Observation. I.e.,
> we could re-use daq:Observation directly"
>
> Following the decision not to re-use DaQ classes directly, the idea is
> - to keep the class dqv:QualityMeasure
> - to declare it equivalent to daq:Observation
> - to declare it as a subclass of qb:Observation (which daq:Observation is)
> [1]
>

+1


> Then we had a brief discussion on whether we should declare explicitly
> dqv:QualityMeasure as a subclass of prov:Entity, which daq:Observation is
> [1].
> I'm not so fond of it. It would make our specification heavier, and
> prov:Entity is not semantically very instructive, anyway.
> I'd have rather some guidelines on representing provenance, following the
> resolutions we make in the other thread. These recommendations could end up
> in infering that some instances of dqv:QualityMeasure are instance of
> prov:Entity (e.g. if they're used as subject of prov:wasGeneratedBy).
> This is how we currently do with dcat:Dataset and dcat:Distribution, as
> shown in our current DQV diagram [2].
>

I am neutral on whether the dqv:QualityMeasure has to be a subclass of
prov:Entity or not. However, I think it is good to show that instances of
dqv:QualityMeasure can be prov:Entity.

If we don't make it a subclass, will we will include the relationships
prov:wasGeneratedBy, prov:generatedAtTime, prov:wasAttributedTo ? I think
it will be good to include them in the diagram similar to dcat:Dataset and
dcat:Distribution and also include them in the examples to encourage their
use. I talked with some colleagues from the W3C Provenance WG and one thing
that they mentioned is that it is good to encourage people to declare the
instances of dqv:QualityMeasure explicitly as prov:Entity in the data so
that without any inference provence tools make use of that data. IMO, this
can be done irrespective of whether the dqv:QualityMeasure is explictly
declared as a subclass of prov:Entity not.

Best Regards,
Nandana