Outgoing

Updated bibstats

We found a bug in the program that computes stastics on fields in WorldCat.  Here is the corrected speadsheet.

The error we found was that column B was supposed to the count of the records with that tag but it ended up being the occurence count.

I apologize for the incorrect data.

--Th

VIAF and RDF

Update: see a more recent post about RDF in VIAF: http://outgoing.typepad.com/outgoing/2010/05/viafs-new-linked-data.html.

We announced Linked Data support for VIAF some time ago, but the actual RDF that we generated was fairly rudimentary FOAF.  Since VIAF is expected to support more than just people (the vision includes corporations, families, geographics, titles, etc.), FOAF seemed fairly limiting, plus the match between FOAF's concepts and typical library authority files is not very good.

The Library of Congress is bringing up LCSH as SKOS, and we have plans here at OCLC to do something similar for the FAST subject vocabulary, so SKOS is an attractive application of RDF for us to use.  Unfortunately SKOS does not have the notion of a person, and lacks some nice RDF classes that FOAF has, such as the ability to specify birth and death dates, or to have a link to a person's publications.

My initial approach was to just bundle up a SKOS and FOAF description of our VIAF personas.  Unfortunately I'm told that our RDF really shouldn't claim that our personas are two different rdf:type's. Andy Houghton here has been helping with this and says that this is not only a bad idea, but that it just doesn't make sense from an OWL/RDF perspective (Keep names separate applies here).

So maybe some of the FOAF properties can be used within SKOS?  Yes, but none of the ones that imply you are describing a person (e.g. you can use foaf:name, but not foaf:publications).  It all makes sense, or at least is consistent, but doesn't help me describe this concept-that-is-a-person. So where are we?  Well, Andy can come up with ways of having multiple real-world-object resources which could be described separately as FOAF and SKOS and then linked, but that seems overly complicated to me.  For right now we'll probably just put out a SKOS concept and people will have to go to the XML object to get more information if they need it.  I'm interested in alternative approaches if anyone has ideas about it.

I can't say I'm happy with much of this.  RDF seems so complicated that it will be very surprising to see it used consistently at all.  I'm reminded of other technologies that needed to be widely deployed, but proved overly complicated, e.g. SGML, and the OSI 7-layer model for communications and ended either morphing into something simpler (SGML to XML) or just ignored (the use of TCP/IP for telecom).

Here's a link to Jane Austen's RDF: http://viaf.org/viaf/39373043.rdf.  Do a 'view-source' on the page to see the XML/RDF.

--Th