Re: Best Practices - Semantic Tagging

On Tue, Mar 5, 2013 at 11:52 AM, Stian Soiland-Reyes
<soiland-reyes@cs.manchester.ac.uk> wrote:
> (We're moving out of the realm of annotations, but I'll keep it on the
> list just for the archive)
>

Thanks. I think I may be unclear where the demarcation is between
annotations and other markup.

> On Tue, Mar 5, 2013 at 1:29 PM, Tim Cook <tim@mlhim.org> wrote:
>
> I would include "How it came to exist, and why" as part of provenance.

Okay, that is what the other metadata is supposed to be used for.
Maybe it needs to be expanded.

> Just be careful about the spelling, xml:id in the xs:complexType, and
> rdf:ID in the rdf:Description. Now don't ask my why they could not
> have the same casing..

Thanks, in actual coding I'll be more careful. :-)


>
> I'm not quite sure about this as it got a bit too specific.. here is my guess:
>
> You keep all the RDF linking complex types vs. purl.bioontology.org in
> a single source. You then automatically embed the relevant statements
> inside the top-level xs:annotation/xs:appinfo of the schema itself
> where those complex types are used (not inside the complexType) .

Yes.

> You'll have to keep your tongue straight on the identifiers if you go
> for this - as the local #ids will be unique depending on the URI of
> the particular schema they are included in  - other than that it
> sounds fine.

I am not sure I understand this.  More below.

>
>>      <rdf:Description rdf:ID="ct-f6c5ea6e-6458-4799-874d-7f3d365d260d">
>>           <rdfs:isDefinedBy
>> rdf:resource="http://purl.bioontology.org/ontology/SNOMEDCT/365761000"/>
>>          < and any other references the modeller wants to create>
>>    </rdf:Description>
>>
>
> So with rdfs:isDefinedBy I would expect to find something about
> #ct-f6c5ea6e-6458-4799-874d-7f3d365d260d on
> http://purl.bioontology.org/ontology/SNOMEDCT/365761000 - but there is
> none.

"isDefinedBy" is an example.  The CCD modeller is a domain expert and
should choose the appropriate rdf or other namespace attribute for the
type of resource they are referencing.

However, the actual complexType name will never be found at those
references.  The referenced resource is intended to be used to
establish the semantics for the UUID based complexType name.
ct-f6c5ea6e-6458-4799-874d-7f3d365d260d on its own, carries zero
semantics.  That is by design, to take the semantics out of the
structural components.

>
> Are you sure rdfs:isDefinedBy is not too strong? I know it's not a
> requirement from rdfs:isDefinedBy, but common usage, at least for
> vocabularies, is that the term would appear in the other end. I would
> recommend checking this on the public-lod list.

I think I addressed this above.

>
>
> It seems more like you want to use a skos relation like
> skos:closeMatch [1] - I would believe you want to say that you mean
> the same concept as
> <http://purl.bioontology.org/ontology/SNOMEDCT/365761000> but that
> your #ct-f6c5ea6e-6458-4799-874d-7f3d365d260d is somewhat differently
> shaped (it's a complex type).

Yes, the SNOMED code 365761000 only points to a location that
describes the concept.  So neither identifier carries any semantics
themselves.


>
>
> As for OA annotations, I think the oa:identifying motivation with an
> oa:SemanticTag body I detailed before should be appropriate. Views
> from the group?

Yes, please.  All input is appreciated.

>
>
> [1] http://www.w3.org/TR/skos-primer/#secmapping
>
>

Thanks.  These are part of the educational materials for knowledge modellers.

--Tim



============================================
Timothy Cook, MSc           +55 21 94711995
MLHIM http://www.mlhim.org
Like Us on FB: https://www.facebook.com/mlhim2
Circle us on G+: http://goo.gl/44EV5
Google Scholar: http://goo.gl/MMZ1o
LinkedIn Profile:http://www.linkedin.com/in/timothywaynecook

Received on Tuesday, 5 March 2013 15:18:45 UTC