It's funny this subject just came up on one of the open-library
discussion lists this week [1]. A whiles ago now Rob Sanderson, Brian
Rhea (University of Liverpool) and I pulled down the LC Classification
Outline pdf files, converted them to text, wrote a python munger to
convert the text into what ended up being a SKOS RDF file. We made the
code available [2] and you can see the resulting SKOS (which needs
some URI work) [3].

It's kind of a work in progress (still). I wanted to get to the point
that the rdf file was leveraged in a little python library (possibly
as a pickled data structure) for easily validating LC numbers and
looking them up in the outline.

I'd be interested in any feedback.

//Ed

[1] http://mail.archive.org/pipermail/ol-lib/2007-September/000069.html
[2] http://inkdroid.org/svn/lcco-skos/trunk/rdfizer/
[3] http://inkdroid.org/tmp/lcco.rdf