Re: How to express nominal dqv:values?

Hi Vladimir,

Thanks for your question.

This problem has been raised a couple of times earlier.

The most notable outcome is that after hesitating we decided to keep to measurements having literal values. 'Nominal' quality assessments are still very much akin to tags. Quite often, if they are based on automatic process, they are based on lower-level quality measurements which do have numerical values. For these cases, QualityAnnotation is not constrained to be used with manual annotations. And it is possible to express derivation links between Annotations and Measurements, as in [1].
And if the nominal quality assessment is manual, then the fit with QualityAnnotation is good too.

Note that your ms:accuracy-medium is quite akin to our :three_stars at [1].

I am not going to say that DQV is going to remain as it is forever, on this specific aspect. DWBP WG is going to produce a W3C Note for DQV, not a Proposed Recommendation. So a good case may influence future re-designs.
But now it will be difficult to change, we would nee much more objective ground than annotations being 'less direct'. The Annotation pattern allows to represent all the information you need. And it is structurally quite similar to the Quality Measurement one...

I'm curious to hear what you think of the issue after this.

Cheers,

Antoine

[1] http://w3c.github.io/dwbp/vocab-dqg.html#expressQualityDerivation

On 07/07/16 12:44, Vladimir Alexiev wrote:
> In the Multisensor project we have some quality ratings of content items.
> These items are called "SIMMO" and have GUID URLs like ms-content:b3f35 (the
> actual URL is much longer).
> The ratings are: low, medium, high, curated (highest).
> Obviously, these are NOMINAL values, not numbers.
>
> I'd like to express them as dqv:QualityMeasurement (more direct), not as
> dqv:QualityAnnotation.
> I did something like this:
>
> ms:accuracy a dqv:Metric;
>    skos:prefLabel "Accuracy"@en;
>    skos:definition "Degree to which SIMMO data correctly represents real
> world facts."@en;
>    dqv:inDimension ldqd:semanticAccuracy;
>    dqv:expectedDataType ms:Accuracy.
>
> ms:Accuracy a owl:Class, skos:ConceptScheme;
>    rdfs:label "Accuracy values"@en.
> ms:accuracy-low a ms:Accuracy, skos:Concept; skos:inScheme ms:Accuracy;
>    skos:prefLabel "Low accuracy"@en.
> ms:accuracy-medium a ms:Accuracy, skos:Concept; skos:inScheme ms:Accuracy;
>    skos:prefLabel "Medium accuracy"@en.
> ms:accuracy-high a ms:Accuracy, skos:Concept; skos:inScheme ms:Accuracy;
>    skos:prefLabel "High accuracy"@en.
> ms:accuracy-curated a ms:Accuracy, skos:Concept; skos:inScheme ms:Accuracy;
>    skos:prefLabel "Manually curated"@en;
>    skos:note "Highest accuracy"@en.
>
> ms-content:b3f35 dqv:hasQualityMeasurement ms-content:b3f35-quality.
> ms-content:b3f35-quality a dqv:QualityMeasurement ;
>     dqv:isMeasurementOf ms:accuracy; dqv:value ms:accuracy-curated.
>
> Problems/questions:
> 1. dqv:expectedDataType rdfs:range xsd:anySimpleType, and as its name
> suggests, it's supposed to point to a literal not resource.
>     But a nominal value is a resource.
> 2. I've "collapsed" owl:Class and skos:ConceptScheme in the sense that:
> -- ms:Accuracy is both owl:Class and skos:ConceptScheme
> -- ms:accuracy-low is both ms:Accuracy and skos:Concept, i.e. has two links
> to ms:Accuracy: rdf:type and skos:inScheme.
> The reason I did it is because it's most natural to model nominal values in
> SKOS; but on the other hand dqv:expectedDataType wants a specific type.
> I think this pattern is not flawed, I think I've seen it for SDMX values.
>
> I think that 1 is a problem in the ontology. 2 is a question of modeling
> best practice.
>
> Can someone help with the above?
> Do you think using nominal values is a significant use case, and thus merits
> a place in the DQV spec?
>
> Thanks in advance!
>
>
>
>

Received on Monday, 25 July 2016 18:03:51 UTC