 |
 |
Whew -- just hit discard on my last message. On Wed, Aug 12, 2009 at 9:07 PM, Karen Coyle<[log in to unmask]> wrote: > then my question is: has B changed? In other words, is B of class X the same > as B of class Y? (Assuming that both B's have the same URI.). "B" (for our purposes we'll say it's "http://example.org/ex/B") can claim it's of as many types as the assertor is willing to predicate (making up words all over this place) as long as none of the classes anywhere assert that they owl:disjointWith (or some similar != assertion) another adopted type. So: <rdf:Description about="http://example.org/ex/B"> <rdf:type resource="http://vocab.org/frbr/core#Manifestation" /> <dct:title>Zen and the Art of the Motorcycle Maintenance</dct:title> <rdf:type resource="http://purl.org/ontology/bibo/Book /> <bibo:isbn10>0553277472</bibo:isbn10> </rdf:Description> Ok -- everything's still in the clear. We've asserted that this resource is a book and that, in FRBR terms, it's also a manifestation. Both of these assertions are true but they're talking about the same resource in different vocabularies -- basically they describe the same thing in different world views: the FRBR model has no knowledge (nor need for knowledge) of bibliographic metadata and vice versa. Now, should you append to this graph something like: <rdf:type resource="http://vocab.org/frbr/core#Text" /> you've run aground. The FRBR schema claims that by being a Text (of course it makes no mention of what exactly that means) it implies also being an Expression but it also defines that <frbr:Expression> <owl:disjointWith> <frbr:Manifestation> (and vice-versa): that is, your resource can't be both an Expression and Manifestation, which makes sense. Now, this doesn't mean that Books and Manifestations are the same thing, it's just that /this/ book also happens to be a manifestation. As far your point about context goes, I think this comes down to trust, credibility and provenance. Even if you define special properties to contain specific parts of your data, there is no way to enforce it. For example, let's say our new RDA vocabulary has: rda:titleProper rda:remainderofTitle and all ILMSes move to an RDA/RDF model (I mean, yes, we're wandering into fantasyland, just bear with me) and begin to store our resources using this as the main data model. Now let's say we have a stash of data we'd like to add to our collection: maybe it's an e-book collection or a set of aggregated OA e-journals, a la DOAJ. The providers of this data are told we need it in our new RDA format and they comply. Let's say, though, that they weren't discriminate enough to distinguish the titleProper from the remainderOfTitle internally but in an effort to comply with our request, put their string in rda:titleProper (it's got to go somewhere, after all) and call it a day. Uncertainty has crept into the mix. After all, there's nothing, technically, stopping me from entering "Zen and the art of motorcycle maintenance: an inquiry into values" all in the 245$a. I think that replacing dct:title with rda:titleProper (rather than declaring that when used in RDA, dct:title should be the proper title) won't drastically help the purity of our data (especially if one of the motivations of RDA and RDF is the promise of externally supplied data) and will have the consequence of being in a vocabulary off the radar for anybody not in a library (and therefore ignored). It's a tough call, though. -Ross.