RE: Supporting arrays of concepts

> From: Leonard Will [mailto:L.Will@willpowerinfo.co.uk] 
> Sent: Tuesday, May 11, 2004 5:21 AM
> Subject: Re: Supporting arrays of concepts
> 
> OK, I agree that it would be useful to have a mechanism for 
> encoding pre-coordinate classification schemes and subject 
> indexing strings, and I do like the idea of treating them in 
> an integrated way that works smoothly with the encoding of 
> thesaurus structures.  It will mean a significant expansion 
> of the project, though. Is it currently within its scope?

I disagree that this would be a significant expansion of the
project.  We have started to do some preliminary mapping of
AAT, LCSH, MeSH, DDC, etc. and for the most part they seem
to map into the SKOS model.  Areas where we are currently
having problems with are notes and node labels.

For notes, I think the problem is easily fixed.  In SKOS,
everything is a skos:scopeNote.  That doesn't fit well with
the vocabularies we are working with.  OK, in a pinch it 
works.  We would prefer that SKOS take the same tack that 
TIF (Thesaurus Interchange Format) took.  There is a Note
class and scopeNote is a subclass of Note.  When you have 
note types that aren't of scope type, you can subclass Note.
Right now, everything has to be sub classed from skos:scopeNote.
That just doesn't feel right and causes some problems with 
Dewey that has between 10 to 20 different note types and 
they aren't all scope notes.  As I said previously, we could 
shoe horn them into skos:scopeNote, but that's not ideal
because we loose the subtle distinctions that the Dewey
editors went to the trouble to make.

Of course, this discussion on node labels, is another area
that is causing us problems.  I'm far from an expert on AAT
but have noted the "guide terms" and the fact that these
"guide terms" can directly have "guide terms" underneath
them.  Dewey also has this same concept in the form of 
centered entries, which can be considered node labels.  The
centered entries can directly have centered entries underneath
them.  Also, in Dewey and as someone else pointed out with AAT 
these "node labels" do have NT/BT relationships.  While in a 
"normal" thesaurus they have only a grouping relationship.
For Dewey, we need to maintain those NT/BT relationships and
I suspect we would need to do the same for AAT.

The current proposal doesn't take this into account and I
really would hate to have two different methods to deal with
the differing strategies.  I think, with a little effort that
both and possible other "node label" strategies could be
accommodated.

After rereading what I wrote here, maybe what is needed is
to separate the relationship information, e.g. NT/BT vs.
grouping.  Also, I think Stella Dextre Clarke mentioned
the ordering problem, e.g. you say that it's ordered
alphabetically -- what doesn't that mean for different
languages?  I think the ordering problem could be easily
handled by attaching audience, e.g. xml:lang, to the
node label array.  Thus you could specify the same
node label array with the elements in differing order
based upon your audience.


Andy.

Andrew Houghton, OCLC Online Computer Library Center, Inc.
http://www.oclc.org/about/
http://www.oclc.org/research/staff/houghton.htm

Received on Tuesday, 11 May 2004 09:32:27 UTC