RE: Library data diagram

Hi Andy,

> OWL constraints essentially apply to the world (or that portion of the
> world that it of interest) - "a 'daughter' can only have one
'mother'".

I would argue that OWL (like any ontology language) is used to create _a
world_ (the universe or domain) one is trying to make assertions about,
rather than just taking _the world_ as the default domain. An OWL
ontology can create any type of universe and relate its individuals in
any possible way (limited by the underlying semantics of the language).

A daughter can have more than one mother in many possible worlds
(including our own, one might add, depending on the
definitions/constraints placed on the hasMother relation).

This is part of the reason why OWL is problematic for schema-type
validation. Rather than rejection a "wrong" assertion, a reasoner would
often just reclassify the individual, e.g., based on domains and ranges
of a relationship. Domains and ranges in OWL are just axioms that
describe the universe (and form the basis of inferences), rather than
constraints to be checked against a set of values.

> Application Profiles constrain/describe what the DC Abstract Model
calls
> Description Sets (what we might have called metadata records in the
past)
> - collections of one or more Descriptions. Application Profiles do not
> describe vocabularies.

The same could be done with OWL(2), but you would likely run into
problems with the open world assumption, amplified by missing
disjointness axioms at some point (see Pete Johnston's post from today,
who also points to resources on how OWL can be used as a constraint
language).

Rules languages (as standardized by RIF) might be able to hit the sweet
spot here: compatible to OWL/RDF and with a place in the semantic web
architecture, but apparently easier for making (and controlling)
assertions not about "a world" but a given "document."

Cheers
Michael

Received on Wednesday, 1 September 2010 18:02:48 UTC