Re: IAU thesaurus in RDF (an update)

From: Ed Shaya <eshaya-at-umd.edu>
Date: Fri, 05 Oct 2007 09:45:34 -0400

Frederic V. Hessman wrote:

>> I don't understand why there are so many TopConcepts, most of which 
>> have  many layers of broader things above it.
>> For instance _21_cm_line is a top concept, yet broader is 
>> radio_recombination_lines, broader is recombination_lines, broader is 
>> emission_lines, broader is specral_lines.  I would perhaps accept 
>> spectral_lines as a TopConcept.

>
> Sorry - I just flushed all of the Concepts right through without
> thinking about it (I'm interested in the tokens, not the
> relationships). It was no problem to throw out Concepts which are
> somebody's NT, now reduced from 2790 (!!) to 991 (!). Are
> "TopConcepts" really so important? I confess I stuck them in only
> because Norman had in his documents and I could just as easily leave
> them out. Presumably, your OWL tools determine top-edness automatically.

This was not the response that I had anticipated. I realized right after I sent the message that this looks just like an Index in the back of a text book with the topLevelConcepts being the top level. For that purpose you might want a couple of hundred items at the top level. I wsa then going to respond, OK, I see. Useful for those (dreaded) humans.

>
> Don't speak OWL, so If someone will suggest a complete and standard OWL
> syntax for the entries, I'd be happy to automatically produce an OWL
> version / include OWL elements as well, as appropriate.

Can you generate a version with the extended SKOS:

skos:broaderPartitive  (PartOf)
skos:broaderGeneric   (subClassOf)
skos:broaderInstantive (InstanceOf, ie the unique physical object like 

ZZ_Ceti, as opposed to ZZ_Ceti_star)
As mentioned here:
http://
isegserv.itd.rl.ac.uk/cvs-public/~checkout~/skos/drafts/appextensions.html ? We don't need narrowerPartitive etc, since that is redundant.

If so, then I can convert to a full blown OWL Ontology with 3 commands in vim. At this point it would be valid but missing the properties between PhysicalObjects and their Measurements. This has to be done manually, I suppose.

Then with a couple of button presses, I can split it into a bunch of reasonably sized namespaced files. I was thinking most skos:broader items, like spectral_lines, could form separate namespaces.

So, we are talking now about having 3 versions, all meant to be official: extended SKOS, SKOS/OWL (just add <owl></owl> and <import skos.owl/>), and OWL (where terms become classes). A script can be written to go from the OWL to SKOS. The other direction loses information.

Ed Received on 2007-10-05Z15:46:18