Re: The SKOS constraint checking use case

In an ideal world, ontology languages (RDFS, OWL, HOL) would be able to 
represent all the logical aspects of a world organization (i.e., an ontology). 
  However, we are not in an ideal world, and all ontology languages are 
representationally inadequate in some way.  (RDFS can't represent local 
ranges.  OWL can't represent three-variable constraints.  HOL can't look into 
the components of its identifiers.)

One way of overcoming these limitations is to have a two-part solution 
consisting of an ontology language (like OWL, for example) and a companion 
language (like SWRL, for example).  The companion language doesn't have the 
same reach as the ontology language and thus has different implementability 
concerns or complexity concerns.  (For example, SWRL doesn't affect OWL 
subsumption reasoning and thus doesn't contribute to the computational 
complexity of subsumption reasoning in OWL.)  The weaker the ontology language 
is the more role there is for companion languages.

It is also possible to use a constraint language (OWL constraints or SPIN or 
whatever) as a companion language, as long as it plays well with the ontology 
language.  (Playing well together is deliberately left undefined here.)  For 
example, constraints could be used to require that start dates are before end 
dates on all events, when known.  OWL constraints and SPIN play well with RDFS 
ontologies, because they naturally tie constraints to classes (and properites, 
at least for OWL constraints).

My comment was whether this use case should be a defining use case.  I had 
thought that it probably shouldn't be, but I now think that it could be.  It 
illustrates a different kind of connection between an ontology and a set of 
constraints.  A set of constraints can be a required adjunct to an ontology, 
associated with an ontology, or completely independent of an ontology.  In the 
latter two cases, which constraint sets are used is under control of some 
external mechanism.


On 11/04/2014 08:56 PM, Holger Knublauch wrote:
> The "Use Case" page
> triggered a response from Peter, and I suggest we continue that topic here in
> email form.
> Peter stated:
> "I would argue that these are no more constraints than much of the rest of
> SKOS. Instead they should be handled by the ontology. Of course, OWL does not
> have the capabilities to state the above condition, so the only recourse is to
> use something like below. The larger question is whether these sorts of
> situations should be considered to be central use cases for the Working
> Group's solution."
> 1) I do not understand what you mean - "they should be handled by the
> ontology". That's exactly what I am trying to do: to have the ontology contain
> enough information to enforce those constraints, so that tools can use a
> generic formalization instead of having to re-code the prose for every
> application.
> 2) You ask whether these scenarios should be considered by the WG, and my
> response is a strong YES - stating such constraints is very much a goal of
> this WG, overcoming the limitations of current languages. We should certainly
> not stop only because OWL doesn't support these scenarios. I wonder how we can
> decide on this question so that we can make progress. Shall we have a vote to
> see what people think?
> Thanks,
> Holger

Received on Tuesday, 25 November 2014 21:37:52 UTC