On Mon, Aug 25, 2008 at 4:30 PM, Martin Malmsten <martin.malmsten_at_kb.se> wrote:
> Let's discuss it on this list for now at least.
Sure, that makes sense. Roy Tennant over in
irc://freenode.net/code4lib was (as usual) prodding me into to
demonstrating something useful about linked-data, and I suggested
perhaps I could write a simple crawler to walk your data set. He
didn't seem to impressed but I continued anyway :-)
30 mins later I had a simplistic 42 line harvester [1]. I let it run
over the weekend (waiting 3 seconds before requests) and it pulled
back 919,190 triples :-) I'm not mentioning this here because it is
some sort of technical feat--quite the opposite. The openness of the
web, rdf's use of URIs, and your data service embracing web
architecture made it possible.
My suggestion to Roy was to imagine a world where library data sets
were linked together, like what you are doing with linking your
authority data with LCSH. A simple crawler could then walk out across
the web, and build a union catalog views automatically using the
collaborative links between systems. Perhaps this isn't the "killer
app" but it feels like it's getting close.
At any rate along the way I noticed a few 500s, which you might want
to look into. I think they all stemmed from the same problem:
http://libris.kb.se/data/bib/5631508
Also, when you link to lcsh.info remember it uses hash URIs to
identify concepts, as you can see in this example:
http://lcsh.info/sh85020816.rdf
And finally when you are creating the links between your subject
authority data and lcsh.info you may want to leverage the SKOS mapping
properties [2] instead of using skos:related.
That's all for now. I'm looking forward to hopefully meeting you at
dc2008 in Berlin!
//Ed
[1] http://inkdroid.org/bzr/linked-data-crawler
[2] http://www.w3.org/TR/skos-reference/#L1309
Received on Tue Aug 26 2008 - 02:58:46 EDT