Re: ConceptScheme as a subset of another?

Hi Mike,

Everyone is on rocky ground here, as these issues have not been investigated much (see [1] for the only thread I can remind of).

Shortly, for the CATCH vocabulary server [2] I was involved in, we used the "multiple inScheme" solution: we create one scheme per "sub-vocabulary", one scheme for the umbrella KOS, and each concept is asserted to belong both to the specific and the umbrella schemes. The positive aspect of this solution is that it allows "basic" data consuming tools to get the info, ie., without requiring them to perform RDFS or OWL inference.

With your option, which is otherwise perfectly valid from a modelling perspective, you have to communicate your new ontology of concepts to data consumer, and require them to "apply" it in inference engines. Of course this drawback disappears if the environment your data is consumed in is controlled, esp. if you're the only one to consume it.

Finally, depending on what you choose in the end (that is, depending on whether your Service and Need concepts belong to a same scheme or to different schemes), your addressesNeed property may be modelled as a sub-property of skos:related or skos:relatedMatch, I think.

Cheers,

Antoine

[1] http://lists.w3.org/Archives/Public/public-esw-thes/2009Jan/0036.html
[2] http://stitch.cs.vu.nl/repository/


> Dear Antoine
> 
> Thanks very much.  That's extremely informative.
> 
> My thinking has moved on a little and I would appreciate any thoughts 
> you have on a slightly different approach. I'm on rocky ground here as 
> my background is more OO and I have very little background knowledge of RDF.
> 
> Essentially subsets are of _like_ things.  So two example ConceptSchemes 
> with subsets might be:
> 
>     * A ConceptScheme of Services with subsets of Fire and of Police
>       services
>     * A ConceptScheme of Needs with subsets of safety needs and of
>       health needs
> 
> Hence I wonder if I should be creating a ex:ServiceClass as an 
> rdfs:subClassOf skos:Concept (and likewise for a NeedClass) and somehow 
> limiting the Service ConceptScheme to holding just ServiceClass concepts 
> (and likewise a Need ConceptScheme to just NeedClass concepts).
> 
> There will be a relationship between Service and Need concepts as 
> defined by a property like ex:addressesNeed.  I'm unsure if such a 
> relationship should be a rdfs:subProperty of a skos mapping or not.
> 
> Lots of issues there.  Any pointers would be welcome.
> 
> Many thanks
> Mike
> 
> 
> On Wed, Jul 15, 2009 at 13:36, Antoine Isaac <aisaac@few.vu.nl 
> <mailto:aisaac@few.vu.nl>> wrote:
> 
>     Dear Mike,
> 
>     Sorry for the delay. To answer your specific questions:
> 
> 
>         I understand that the skos:inScheme property can be used to say
>         a concept is in a particular ConceptScheme and this can be used
>         many times to say the same concept is in many schemes.
> 
> 
> 
>     Exact.
> 
> 
>         I can't say that one ConceptScheme is in another can I?
> 
> 
> 
>     You cannot say this, at least using the SKOS model or another
>     standard Semantic Web vocabulary. See [1].
>     We have discussed that in the WG, but the requirement was not clear
>     enough at the time.
>     The only standard way is really to have your concept belonging to
>     several schemes at a same time.
> 
> 
>         Is there any sensible way I can support these subsets?
> 
> 
> 
>     Now, you could still create your own property to represent scheme
>     inclusion, and equip it with semantics that help you implement the
>     multiple inScheme statement trick.
>     With OWL 2, this is doable:
>     1. you can introduce, say, an ex:subScheme property
>     2. you can define a property chain axiom [2] that combines it with
>     skos:inScheme: [] rdfs:subPropertyOf skos:inScheme;
>     owl:propertyChainAxiom (skos:inScheme ex:subScheme ).
> 
>     With this, if A is a sub-scheme of B, whenever you have c1
>     skos:inScheme A, then you would also have c1 skos:inScheme B if you
>     use an appropriate reasoner.
>     But of course this solution is not guaranteed to work well when
>     others consume data that you wouldn't have "completed" youself
>     beforehand, as it would rely on an adhoc property...
> 
>     Cheers,
> 
>     Antoine
> 
>     [1] http://www.w3.org/TR/skos-primer/#secextension
>     [2] http://www.w3.org/TR/owl2-new-features/#F8:_Property_Chain_Inclusion
> 
> 
>          
>          
>         ------------------------------------------------------------------------
> 
>         *Van:* public-esw-thes-request@w3.org
>         <mailto:public-esw-thes-request@w3.org>
>         [mailto:public-esw-thes-request@w3.org
>         <mailto:public-esw-thes-request@w3.org>] *Namens *Mike Thacker
>         *Verzonden:* dinsdag 30 juni 2009 18:25
>         *Aan:* SKOS
>         *Onderwerp:* ConceptScheme as a subset of another?
> 
>          
>         Hello
> 
>         I'm getting my head around SKOS with a view to seeing how well
>         we can represent UK public sector vocabularies hosted by
>         esd.org.uk <http://esd.org.uk> <http://esd.org.uk> in SKOS and
>         derive online viewers and other human and machine readable
>         resources from the RDF.
> 
>         My initial question relates to whether or not SKOS supports what
>         esd knows as "subsets".  For example these two online viewers:
> 
>            * Subject vocabulary:
>         http://www.esd.org.uk/standards/ipsv/viewer/
>            * Service vocabulary:
>         http://www.esd.org.uk/standards/lgsl/viewer/
> 
>         both feature drop down lists towards the top left of the page.
>          >From the drop-down list you can select a named "subset" which
>         contains some of the concepts from the main list (eg you can
>         limit UK local government services to those delivered in Scotland).
> 
>         I understand that the skos:inScheme property can be used to say
>         a concept is in a particular ConceptScheme and this can be used
>         many times to say the same concept is in many schemes.  I can't
>         say that one ConceptScheme is in another can I?
> 
>         Is there any sensible way I can support these subsets?
> 
>         Thanks in advance
>         Mike
> 
> 
>

Received on Tuesday, 21 July 2009 08:12:37 UTC