W3C home > Mailing lists > Public > public-lod@w3.org > February 2013

Re: How can I express containment/composition?

From: Bernard Vatant <bernard.vatant@mondeca.com>
Date: Thu, 21 Feb 2013 16:31:04 +0100
Message-ID: <CAK4ZFVGQo=kiy+tzYFoArot6D+Cuc=V3R8ELrzJSqtrXd0zzmg@mail.gmail.com>
To: Barry Norton <barry.norton@ontotext.com>
Cc: public-lod@w3.org
Hi all

(with my Geonames ontology editor helmet on)

2013/2/21 Barry Norton <barry.norton@ontotext.com>

>
> I agree that one should expect (some) geographical containment(s) to
> represent general partonomy; I guess geonames doesn't because there is no
> canonical property for partonomy.
>

Well, geography is a tricky domain ... gn:parentFeature is the most generic
hierarchical property between gn:Feature, but it's not a partonomy stricto
sensu, because it covers both physical and administrative features. And its
transitivity entails for example that Guadeloupe is both child of Europe
through its administrative parent which is France, and child of Northern
America by its geographical location.
In a strict partonomy, seems to me that having two ancestors which are
physically disjoint should be forbidden.

gn:parentFeature   rdfs:subPropertyOf   dcterms:hasPart
would make sense since the latter has also a quite loose definition. I put
it on the to-consider list for the next publication.
"A related resource that is included either physically or logically in the
described resource."

http://www.w3.org/2001/sw/BestPractices/OEP/SimplePartWhole/part.owl#partOfhas
probably a stronger semantics.

The double parenthood above would certainly break the results of population
using SPARQL queries such as proposed by Matteo if you count by continent.
People from Guadeloupe will be counted both as European and American ...

E.g., Geonames has:
>
>  :parentFeature     a owl:ObjectProperty,
>                 owl:TransitiveProperty;
>          rdfs:comment "A feature parent of the current one, in either administrative or physical subdivision."@en;
>          rdfs:domain :Feature;
>          rdfs:label "parent feature"@en;
>          rdfs:range :Feature .
>
>  :parentADM1     a owl:ObjectProperty;
>          rdfs:domain :Feature;
>          rdfs:label "level 1 administrative parent"@en;
>          rdfs:range  [
>              a owl:Restriction;
>              owl:hasValue <#A.ADM1>;
>              owl:onProperty :featureCode ];
>          rdfs:subPropertyOf :parentFeature .
>
>
> But no
>
>
> :parentFeature rdfs:subPropertyOf http://www.w3.org/2001/sw/BestPractices/OEP/SimplePartWhole/part.owl#partOf.
>
>
> Or
>
>
> :parentFeature rdfs:subPropertyOf dcterms:hasPart.
>
>
> To be honest though, I don't like your example as being a citizen of a
> country *doesn't* imply a partonomy relationship to me (neither might
> some kinds of geographic containment - I'm thinking Lesotho-South Africa).
>
> Barry
>
>
>
> On 21/02/13 14:10, Frans Knibbe | Geodan wrote:
>
> Barry and Matteo, thank you for pointing me to the GeoNames Ontology.
> Geographical containment can also be found in GeoSPARQL (
> http://schemas.opengis.net/geosparql/1.0/geosparql_vocab_all.rdf):
> sfContains.
>
> I had the feeling that what I primarily needed was the logical concept of
> containment/composition, because that would allow reasoning on the part of
> the data consumer. But I guess it would be best to specify both logical AND
> geographical containment. As far as I can tell, the geographical
> containment in GeoSPARQL and GeoNames does not imply logical containment.
> But perhaps I am overestimating the power of dcterms:hasPart?
>
> I was thinking about an example. Let's say the following is known:
>
> 1) A country consists of provinces
> 2) For each country, the complete set of provinces is available
> 3) For each province the number of inhabitants is available
>
> Could a machine answer the question "Which country has the highest number
> of inhabitants?" without help from a human?
>
> Regards,
> Frans
>
>
>
> On 21-2-2013 14:10, Matteo Casu wrote:
>
> You could also check the GeoNames ontology, which considers administrative
> subdivisions: http://www.geonames.org/ontology/documentation.html
> E.G.: in the USA, level 1 administrative subdivisions are States. In
> Italy, they are Regions.
>
> It is a minor change of perspective with respect to yours.
>
>
> Il giorno 21/feb/2013, alle ore 14:01, Frans Knibbe | Geodan
> <frans.knibbe@geodan.nl> <frans.knibbe@geodan.nl> ha scritto:
>
> Thank you Martynas, that seems to be just what I was looking for!
>
> Frans
>
> On 21-2-2013 13:54, Martynas Jusevičius wrote:
>
> Hey Frans,
>
> Dublin Core Terms has some general properties for this:
> dct:hasPart http://dublincore.org/documents/dcmi-terms/#terms-hasPart
> dct:isPartOf http://dublincore.org/documents/dcmi-terms/#terms-isPartOf
>
> Martynas
> graphity.org
>
> On Thu, Feb 21, 2013 at 2:47 PM, Frans Knibbe | Geodan
> <frans.knibbe@geodan.nl> <frans.knibbe@geodan.nl> wrote:
>
> Hello,
>
> I would like to express a composition relationship. Something like:
> A Country consist of Provinces
> A Province consists of Municipalities
>
> I thought this should be straightforward because this is a common and
> logical kind of relationship, but I could not find a vocabulary which
> allows
> be to make this kind of statement. Perhaps I am bad at searching, or maybe
> I
> did not use the right words.
>
> I did find this document:
> http://www.w3.org/2001/sw/BestPractices/OEP/SimplePartWhole/ ("Simple
> part-whole relations in OWL Ontologies"). It explains that OWL has no
> direct
> support for this kind of relationship and it goes on to give examples on
> how
> one can create ontologies that do support the relationship in one way or
> the
> other.
>
> Is there a ready to use ontology/vocabulary out there that can help me
> express containment/composition?
>
> Thanks in advance,
> Frans
>
>
-- 
*Bernard Vatant
*
Vocabularies & Data Engineering
Tel :  + 33 (0)9 71 48 84 59
Skype : bernard.vatant
Blog : the wheel and the hub <http://blog.hubjects.com/>
--------------------------------------------------------
*Mondeca**          **                   *
3 cité Nollez 75018 Paris, France
www.mondeca.com
Follow us on Twitter : @mondecanews <http://twitter.com/#%21/mondecanews>
----------------------------------------------------------

Meet us at Documation <http://www.documation.fr/> in Paris, March 20-21
Received on Thursday, 21 February 2013 15:32:03 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 31 March 2013 14:24:46 UTC