[Metware-devel] SKOS validation

[Egon W. <ego...@gm...>]

Hi all,

pending replying the Steffen's recent proposals, I worked on the
current SKOS versions of the ArMet and MetWare ontologies, and managed
to get Bioclipse2 to validate the SKOS, and added the involved
JavaScripts to BigMet/onto/validate/...

During this process, I cleaned up the SKOS of both, and added axioms,
so that skos:Concept's from the ArMet skos:ConceptScheme, are now also
rdf:type ArmetConcept, and likewise for MetwareConcept.

I found that this was needed because the current set up uses
skos:broader to link MetWare derivatives to their ArMet ancestors.
There is rumour of SKOS-Mapping to do this, but have not found
something definite... I need this to test that all Armet concepts have
equivalents in MetWare... (this will later be used for the MSI
ontology too, which extends Armet, and we will want to be compatible
and thus extend the MSI onto too)

Anyway, the current sources use OWL axioms and only typing the
skos:ConceptScheme "&armet;armet" as armet:ArmetConceptScheme too...
the armet:ArmetConcept rdf:type-ing follows from the axioms when using
a reasoning SPARQL engine like Pellet, as I do in the RDF feature of
Bioclipse2.

So, what does this mean practically... I have now written the first
few unit tests for the SKOS files, testing these assumptions:

* all skos:Concept in armet.skos must be a ArmetConcept
* all skos:Collection in armet.skos must be a ArmetCollection

Test results:

RDFStore: 1068 triples
Model is valid.
Collections not in the Armet scheme: 0
Concepts not in the Armet scheme: 0

* all skos:Concept in metware.skos must be a MetwareConcept
* all skos:Collection in metware.skos must be a MetwareCollection

Test results:

RDFStore: 2718 triples
Model is valid.
Collections not in the Metware scheme: 0
Concepts not in the Metware scheme: 19
[[http://metware.sf.net/onto/DataPreprocessingAnalysis, null],
[http://metware.sf.net/onto/Organism, null], [http://metware.sf.net/onto/Run,
null], [http://metware.sf.net/onto/User, null],
[http://metware.sf.net/onto/Observation, null],
[http://metware.sf.net/onto/Protocol, null],
[http://metware.sf.net/onto/Machine, null],
[http://metware.sf.net/onto/Metabolite, null],
[http://metware.sf.net/onto/Project, null],
[http://metware.sf.net/onto/AnalysisMaterial, null],
[http://metware.sf.net/onto/Experiments, null],
[http://metware.sf.net/onto/Tissue, null], [armet:mchSetupRef, null],
[armet:runID, null], [http://metware.sf.net/onto/Institute, null],
[armet:PreprationMethod, null], [http://metware.sf.net/onto/Experiment, null],
[armet:archiveRef, null], [http://metware.sf.net/onto/Organs, null]]

Clearly, I need to clean up the metware.skos a bit more. Three
principle problems:

* incorrect namespacing: http://metware.sf.net/onto/Protocol should be
http://metware.sf.net/onto/#Protocol
* there are a few put in the armet namespace in the metware.skos
* 19 concepts are not assigned to a ConceptScheme yet

Will look at this tomorrow...

Egon

-- 
Post-doc @ Uppsala University
http://chem-bla-ics.blogspot.com/