Re: SKOS Instance V Classes

On 2024-01-31 02:58, Antoine Isaac wrote:
> > > If I may add, wrt OWL2, indeed it was formally finalized after we had to wrap up the SKOS specs. There are more details about SKOS and OWL in sections 5.2 and 5.4 of this wrap-up paper: https://doi.org/10.1016/j.websem.2013.05.001
> > > 
> > > Also, maybe a relevant blog post from a user back then is https://www.mkbergman.com/944/skos-now-interoperates-with-owl-2/
> > > 
> > > Finally I could note that Renato's example can also apply to properties, not only to classes: the Library of Congress has a vocabulary of concepts that are dually defined as (OWL) properties:

Thank you, Renato, for initiating this interesting exchange.

We are testing an approach that embraces the notion of punning property-like SKOS concepts as RDF properties in a project at the National Agricultural Library, NALT for the Machine Age.

The project aims at supporting the integration of non-"Semantic" agricultural data - data in spreadsheets about harvests, soil, weather - on the basis of RDF.

The challenge is to help research scientists achieve a working consensus about their entities of interest, and about the properties of those entities; to capture that consensus in boxologies (box-and-arrow models); and, by adding constraints on properties and values, to turn those boxologies into data shapes that can then serve as normalization targets for data integration.

It is in our experience hard to explain to research scientists the semantic distinction between a SKOS concept for, say, "micronaire" (nalt:123249) - defined as "a combined measure of cotton fiber fineness and maturity" - and an RDF property with essentially the same definition. We figure that machines will have no trouble seeing that nalt:123249, if used in the predicate position of a triple, is being used as a property.

We want to capture and record in NALT itself the sets of concept/properties (and possibly concept/classes) that have previously been used in boxologies and data shapes. Property and class vocabularies tend to be smaller than, and packaged separately from, SKOS concept schemes, so this looks like a good use for SKOS collections. Specified orthogonally to the hierarchy of SKOS concepts, collections can serve as "starter sets" for creators of future boxologies and data shapes.

I recently described this work in a short talk at an NKOS workshop [1,2].

We are writing a paper to describe this in more detail. If anyone else out there is working along these lines, please do get in touch so we can learn from and point to your work.

Tom

[1] https://www.youtube.com/watch?v=NyRlk29n3bk&t=12023s
[2] https://nkos.dublincore.org/2023NKOSworkshop/NKOS2023.html

-- 
Tom Baker <tom@tombaker.org>

Received on Tuesday, 6 February 2024 11:05:52 UTC