RE: ssn: issue-72 inverse properties in sosa-core

Ø  I'd add rdfs:domain, rdfs:range, and owl:inverseOf in SOSA.

Ø  … chema:domainIncludes and schema:rangeIncludes may be commented out some day

This was discussed at length in a couple of the meetings.

The general sense of the discussion was was that

-          rdfs:domain and rdfs:range entailments are strong and might prevent use of SOSA properties by people who don’t want their things to be implicitly sub-classes of something – AFAIK there has been a general move away from these global constraints in pragmatic ontology engineering anyway, in order to make re-use of individual terms easier.

o   OTOH domainIncludes/rangeIncludes are ‘permissive’, hints if you like, which have no formal entailments, but have been found useful by schema.org

-          owl:inverseOf (especially in the absence of global domain/range axioms) is essentially just formalizing the documentation – we want to say ‘this property is the inverse of that one’, and an OWL predicate is available that already does that, so why not use it

o   it will be ignored by non-ontologists anyway.
So your proposal looks exactly backwards from the decisions already made, in terms of SOSA at least.

Remember – the primary audience for SOSA is not ontologists. The goal is that it is not inconsistent with a more formal system that can be built on top of it, but SOSA is deliberately incomplete in both scope and axiomatization. More rigorous ontology engineering is done in vertical extensions, including SSN.

Simon


From: Maxime Lefrançois [mailto:maxime.lefrancois@emse.fr]
Sent: Friday, 10 February, 2017 05:20
To: Kerry Taylor <kerry.taylor@anu.edu.au>; janowicz@ucsb.edu; Raúl García Castro <rgarcia@fi.upm.es>; public-sdw-wg@w3.org
Subject: Re: ssn: issue-72 inverse properties in sosa-core

Hi Kerry,

I thought it made sense to keep to rdf alone.
 I understood the sentiment favoured instead some kind of “simple” OWL.

I believe so, it makes then a bit strange to try to "prevent" the use of OWL reasoners with SOSA. I'm not very fond of schema at all, and if you ask, I'd add rdfs:domain, rdfs:range, and owl:inverseOf in SOSA.

But I think I understood that schema:domainIncludes and schema:rangeIncludes may be commented out some day, as is the case in QB4ST ?

 I, for one, never really understood  what “simple” means for sosa, but I suppose for some people it means just exactly what is in sosa now. And  just now we agreed that owl:AnnotationProperty  can appear in sosa, but I guess that that is “simpler” than owl:inverseOf.
yes, owl:inverseOf is more complex than the annotation property schema:inverseOf.

Anyway – I think, from the arguments at the time, the schema.org<http://schema.org> solution you propose would be interpreted  as the same as “documentation” (B),   and therefore has been decided already, although I don’t recall your (C) as coming up  at the time.

(C) would have been the option I would have proposed if I was following the discussion at that point. It consists in using schema:inverseOf the exact same way we use schema:domainIncludes and schema:rangeIncludes, i.e. for example:

schema:domainIncludes a owl:AnnotationProperty .
schema:rangeIncludes a owl:AnnotationProperty .
schema:inverseOf a owl:AnnotationProperty .

sosa:hasFeatureOfInterest a owl:ObjectProperty ;
  schema:domainIncludes sosa:Observation ;
  schema:rangeIncludes sosa:FeatureOfInterest ;
  schema:rangeIncludes sosa:Sample ;
  schema:inverseOf sosa:isFeatureOfInterestOf .


So +0 from me.

Btw – the earlier question about “meta:” – mea culpa – but I stand corrected.  Great to have fresh eyes on this.

no worries, but I just saw that the following documents also use it:
 - ssn/rdf/sam.ttl
 - ssn/rdf/om.ttl
 - qb4st/ontology/qb4st.ttl

Kind regards,
Maxime