« VIAF article | Main | Updated bibstats »


Bruce D'Arcus

"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."

The issues you are grappling with have nothing to do with RDF per se; it's about fitting your local data into a global information space. This includes fitting incredibly idiosyncratic library data traditions into more general vocabularies. I really commend you for trying to do this, but think it's important to properly diagnose the problem.

Ryan Shaw

I would recommend creating your own classes and properties that fit the semantics of authority records (as you understand them), and let others worry about how they will map them to FOAF or SKOS. (This is the approach Freebase and DBpedia have taken.) I don't think SKOS is such a great fit for authority records, and as you point out, neither is FOAF. People will tell you it's best practice to reuse vocabulary, and that's true, but when you have a data set as large and important (and idiosyncratic) as yours, I think it's better not to shove it into a poorly-fitting data model. The "use a SKOS concept and people will have to go to the XML object to get more information if they need it" approach is the worst of all worlds, because now you're forcing people to use two different toolsets and two different models (graphs and trees) to get at your data, when the whole point is to make your data easily consumable. Far better to either make all the data available as RDF using a VIAF-specific vocabulary, or just forget RDF and use Atom or something. (Or even better: do both.)


I agree with what Bruce said about RDF not being the problem. The problem lies in RDF forcing you to think about data modeling issues instead of just sweeping them under the carpet. Mark Pilgirm's comments http://www.xml.com/pub/a/2003/08/20/dive.html on Atom and RDF from 2003 still seem relevant today to me.

I guess I probably missed it on public-esw-thes or foaf-dev, but I'd be interested to see the discussion of why you can't use FOAF for people, and GeoNames Ontology for places, etc. Also I'm not sure I see how the "Keep Names Separate" applies to instance data. I also agree with your assessment that minting different URIs for the FOAF concept and SKOS concept, and linking them together seems overly complicated. I still don't see what's wrong with typing a resource as skos:Concept and foaf:Person honestly.

Thanks for blogging about this Thom. While I agree with Ryan's comments too I think that a good faith effort needs to be made to reuse vocabularies...and you are clearly doing that. I've heard Tom Baker and Antoine Isaac talk about starting up a semweb for libraries group either at the W3C or somewhere else. I definitely think having a telecon between people would help a lot in coordinating these things.

Jonathan Rochkind

I share some skepticism about RDF. Although I'm sure you could model your stuff as RDF effectively; if it requires creating your own vocabularies that nobody else uses anyway, is it worth the trouble? Some will strenuously argue 'yes'. I'm not sure.

But the important thing is to make the data available in machine readable format, and with URIs for each concept/entity involved. If you can do this effectively in RDF, great. If your winding up spending so much time trying to figure out how to do it in RDF that you wonder if it's worth it, and put it up in some kind of XML (I can't think of anything but some kind of XML that will be practically useful for the consumer, if not RDF) -- I personally will still be satisfied, and I suspect most actual potential users of the data will be too, even if the RDF theorists/purists will complain.

Most of the promise of RDF (as opposed to some kind of XML, even custom XML schema) is, from my perspective, mostly just _promise_, not currently delivered. Now the benefit of the promise is great enough that, all things being equal, if it's just as easy to do it in RDF as something else, sure, do it in RDF, support the RDF experiment.

But, IMO, if it doing it in RDF is going to mean not doing it at all (or delaying it significantly) as you try to figure out how to do it -- just start out with non-RDF XML, the most important thing is making the data available in an easy to use format.

Karen Coyle

SKOS was developed for thesauri, not for authority data. It barely, if at all, works for LC subject authorities, and really should not be used for name authorities, IMO. We need to create SAOS - Simple Authority Organization System - or maybe even CAOS - Complex Authority Organization System. It would be interesting to work on this in conjunction with the FRAD model. I have a lot of questions about that model (such as the treatment of preferred names and identities as entities), and this might allow us to work through some issues about how the model might work in real systems.

The comments to this entry are closed.

My Photo

April 2018

Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30