W3C home > Mailing lists > Public > public-dwbp-wg@w3.org > March 2016

Re: DQV, ISO 19115/19157 and GeoDCAT-AP - Issue-202 - representing conformance

From: Antoine Isaac <aisaac@few.vu.nl>
Date: Thu, 10 Mar 2016 00:01:39 +0100
Message-ID: <56E0AB53.5060406@few.vu.nl>
To: Andrea Perego <andrea.perego@jrc.ec.europa.eu>
CC: "public-dwbp-wg@w3.org" <public-dwbp-wg@w3.org>
Hi Andrea,

No problem with long mails, as long as they give the opportunity for productive discussions!
And I believe this is the case here. And like, as in *hugely* productive.

First I'm going to add a reference Issue-202 so that our tracker can keep tracking:
https://www.w3.org/2013/dwbp/track/issues/202

Second I'm noting this becomes marginally related to conformance levels. It's more about conformance statements in general. I guess we'll come back to levels later...

Now I'm going to try and answer your points:


On 3/8/16 2:31 PM, Andrea Perego wrote:
> Hi, Antoine.
>
[snip]
>> - we are about to add in DQV examples regarding quality policies (draft
>> at [1]). Do you think this is closely related to the issues you raised
>> here? Should we unify the patterns? At this stage I'd rather avoid the
>> extra work, but I do have to check with you.
>>
>> [1]
>> https://www.w3.org/community/odrl/wiki/W3C_Data_on_the_Web_Best_Practices_-_Data_Quality_Policy
>
> Yep - I think it's definitely related to the scenarios I was referring to.
>
> If I correctly understand, the idea in [1] is to use ODRL to provide a formal description of a quality policy (in this case, an SLA).


Yes.


> So, the computed quality measurements can be matched against what said in the quality policy (_:ex1) to determine whether the relevant resource (:myDatasetDistribution) complies with it (:myDatasetDistribution dct:conformsTo _:ex1).


We had not touched the problem yet. But of course it should be possible to (numerical) quality measures correspond to tests for checking conformance with SLAs or other policies. Especially when the quality dimensions are relevant for SLAs.


>
> Quality measurements and conformance statements could be both seen as "observations" (in the broader sense), so aligning their representation shouldn't be a problem, conceptually.


Agreed. This would be desirable, even.

NB: before continuing, I'm going to assume familiarity with the current DQV patterns:
http://w3c.github.io/dwbp/vocab-dqg.html#model


> However, I don't know if the current version of DQV can already support this.In particular, two points are unclear to me:
>
> 1. Can dqv:QualityMeasure be used to express a conformance statement?
>
> This would require, e.g., expressing the quality measure "value" (dqv:value) with code lists as those used in ISO, EARL and INSPIRE. However, in the examples currently included in DQV, dqv:value is used only with literals.


Yes. And I think the idea (which is also the one of daQ) is that we keep Measurements to numerical values, generally. This also gives some benefits related to interoperability with RDF Data Cubes.

If there's a quality observation that is more 'semantic' (e.g. with a controlled vocabulary) I'd expect it is rather done with QualityAnnotations, not QualityMeasurement
That being said I believe we can have some alignments (see below).



>
> On this issue, it is worth mentioning the ISO approach:
> https://geo-ide.noaa.gov/wiki/index.php?title=ISO_Data_Quality
> As you can see from the UML diagram, ISO has the notion of data quality report (DQ_Element), where the result can be either a quantitative result (DQ_QuantitativeResult) or a conformance result (DQ_ConformanceResult). As far as I can see, dqv:value models just the former (quantitative result).


Right.

In DQV now the final conformance statement could be either expressed as dcterms:conformsTo or QualityAnnotations.
I think what we originally meant was to reflect differences of 'authoritativeness' for the conformance statement. We (at least I, at the time) felt that publishers of datasets should be allowed to express their conformance statements in a more direct way than agencies or just any user (and here the difference of authority could be more obvious)  'tagging' the dataset as being conformant.

But in the light of the discussion we had together these last days, I'm now tempted to accept that there is some mapping between both, as you do with GeoDCAT-AP. The difference would just one of granularity...

>
>
> 2. Which is the relationship between dqv:QualityMeasure and dqv:QualityPolicy?
>
> I don't know if this is already supported, but I think that an explicit relation between these entities will "close the circle" between two relevant branches of DQV, making it possible to align the representation of quality measures and conformance statements.
> About "closing the circle", but on a different part of DQV:
> Is there any relationship between dqv:QualityAnnotation, dqv:QualityMeasure and dqv:QualityPolicy? This would allow the association of a quality annotation (e.g., in the form of a quality certificate) with the computed quality measures, demonstrating conformance with a given quality standard (e.g., the ODI certificate).
>
>

In fact the loop may be closed by the other discussion we have about Issue-222 on 'derived metrics', with relations like 'dependsOn', 'isUsedToObtain', 'wasInfluenceBy'... [1,2]
If we generalized the relations discussed there to be applicable to other classes than Metric or QualityMeasurement, we could express that a QualityAnnotation or QualityCertificate is based on a QualityMeasure.

I think this could go a long way meeting your requirement. What do you think of this?

If you agree then we would modify the scope of Issue-222, to make sure the outcome of that one can be used for closing the loop!

Cheers,

Antoine

[1] https://lists.w3.org/Archives/Public/public-dwbp-wg/2016Feb/0045.html
[2] https://lists.w3.org/Archives/Public/public-dwbp-wg/2016Mar/0026.html
Received on Wednesday, 9 March 2016 23:02:11 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 9 March 2016 23:02:12 UTC