Re: DQV Comments

Hi Antoine,

I think we could look at constructing an architecture for DQ that focuses
on corroboration and validation, in which all sources of Data on the Web
should at least be doubted without x number of corroboration sources, some
lineage validation, and some data decay formula.  The architecture could
focus on the technical characteristics of linked data corroboration and
lineage validation and we could insert methods for using the architecture
into the BP document.

Happy to discuss on the next DQV call.  When is it?

Best Regards,


Motto: "Do First, Think, Do it Again"

| From:      |
  |Antoine Isaac <>                                                                                                                  |
| To:        |
  |"" <>                                                                                                   |
| Date:      |
  |06/19/2015 08:12 AM                                                                                                                               |
| Subject:   |
  |Re: DQV Comments                                                                                                                                  |

Hi Steve,

You're reading it fully right.

The first answer is that we need to have a very general framework, so that
everyone can create data that represent their own metrics or subjective
assessments, which they think should be done. It is very difficult to
anticipate the diversity of needs, and we don't want to create a
straitjacket that would prevent too many people from extending our DQV in
the way they need it.

The second answer is that such an abstract framework is not very
satisfactory for us editors, too. This is why we've created a section on
example usage [1], if just to give ideas to readers on how to use the
generic framework. This is still work in progress, as you can see.

The problem is that we don't have much input from the DWBP use cases and
best practices. They too mention quality in a very abstract way, as we've
found in our analysis [2,3]. I've repeatedly asked for more input (e.g.
during the last F2F), but it seems really not easy for the group to do it.

So if we want to add stuff now, we have to look outside the current UCs and
BPs, as I think our investigations on completeness has showed [4].
In fact we have some leads. For example, a colleague from VU, Davide
Ceolin, works on expressing ISO quality criteria for datasets as RDF.
But doing more requires time, which the editors alone don't have. Actually
for the last F2F we had even started to create a questionnaire [5] but none
of us could push it further.

And in any case, it will not provide a mandate for including properties and
classes for specialized metrics or subjective assessments (annotations) in
DQV itself, if this is what you're after.
Fitting specialized classes and properties as part of the standard is
possible only if we find evidence of general need for them, preferably
based on the material the group has gathered.

But of course if you have ideas (and it seems you have some) to extend the
current material we must start with, it's more than welcome!

Kind regards,



On 6/15/15 8:24 PM, Steven Adler wrote:
> Just reading through the DQV and apologize if my comments are uninformed.
>From my experience with Data Quality, one can make objective observations
about Data Quality based on the age of the data, in which one assigns
various decay factors and calculates age(decay factor), it's completeness
in which a percentage can be applied.  One can also apply subjective
assessments by comparing one data set to other sources.  One can even
assert that data without comparative sources can't be trusted - no unbribed
journalist would ever publics a story without corroborated sources...
> But when I read the DQV I don't really get these points and the whole
thing feels very abstract.  Am I reading it wrong?
> Best Regards,
> Steve
> Motto: "Do First, Think, Do it Again"
> Inactive hide details for Laufer ---06/11/2015 09:26:29 PM---Hi Antoine,
Ok. I think is good to have a self-contained document.Laufer ---06/11/2015
09:26:29 PM---Hi Antoine, Ok. I think is good to have a self-contained

> Hi Antoine,
> Ok. I think is good to have a self-contained document.
> Best,
> Laufer
> Em quinta-feira, 11 de junho de 2015, Antoine Isaac < <>> escreveu:
>     Hi Laufer,
>     Thanks for the comment!
>     We've just followed existing practice in DCAT. Ie. DCAT re-uses the
skos:Concept class, and still "re-defines" it in the DCAT reference doc
>     I guess other 'vocabulary documentation schools' would not reproduce
the external info. But I do like the idea of having a self-contained
document, at least as long as the effort is not huge.
>     And in the case of DQV and DAQ there's another point: as pointed
explicitly (as an ISSUE) in the DQV draft, we may end up have to re-declare
the DAQ constructs as DQV (or even DCAT) ones, later. In that case it will
have been a smart move to have the doc self-contained, earlier than later.
>     Kind regards,
>     Antoine
>     [1] _
>     On 6/11/15 8:01 PM, Laufer wrote:
>         Hi, Antoine, Christophe, Riccardo,
>         First of all, thank your for your efforts in DQV.
>         I have a question about the DQV Data model  (Fig.1):
>         Considering that dqv:QualityMeasure is a subclass of
daq:Observation, and that
>         the relations beetwen daq:Observation, qb:Observation, daqMetric,
daq:Dimension, daq:Category are defined in
>         it is necessary to have qb:Observation, daqMetric, daq:Dimension,
daq:Category explicitly defined in DQV Data Model?
>         Thank you.
>         Best Regards,
>         Laufer
>         --
>         .  .  .  .. .  .
>         .        .   . ..
>         .     ..       .
> --
> .  .  .  .. .  .
> .        .   . ..
> .     ..       .

Received on Monday, 22 June 2015 14:20:20 UTC