Re: Question about MARCXML to Models transformation

On Tue, Mar 8, 2011 at 11:01 AM, Karen Coyle <kcoyle@kcoyle.net> wrote:

>
> And I thank you, Ross, for that and for your clarity in this discussion.
>

Hey, whenever you need somebody to come in and punt on the hard problems,
I'm your man.


> I still think it should be possible to model bibliographic data using the
> FRBR or RDA attributes but without the division into WEMI for Group 1 data.
> (e.g. the RDA generalized attributes) If you have a work title, well, you
> have a work title. You don't need a work entity to have a work title. (I
> realize fully that I may be wrong about this, but this is what I *want* to
> be right.) The predicates and objects should stand on their own without the
> separation (and separate identifiers) that WEMI entities require. Where this
> falls apart is in the WEMI-to-WEMI relationships, but I just figure there
> must be a way to solve that as a practical problem so that we don't have to
> shoe-horn everything into WEMI as separate entities.
>

I absolutely agree with this.  One of the really nice things about SKOS
(imo) is that it only defines domains and ranges on the properties that are
kos-specific (broader/narrower/inScheme/etc.) and not to the things that are
generally useful even outside the context of SKOS (prefLabel, altLabel,
notes, etc.).   This way we now have a consistent means of defining a
preferred and alternate labels, for any resource, without entailing that the
resource is also a concept.

The same definitely needs to apply to (some) RDA terms.  We should be able
to model resources in BIBO or DC Terms (or whatever), enhance with the more
domain-specific properties RDA brings to the table, without muddying things
up semantically by dragging in FRBR along with it.

As far as the WEMI-to-WEMI relationships (and by this, I think you mean
relationships like adaptations, translations, derivative works, etc.,
right?), you're right, this gets a bit more complicated.  I still think you
can abstract a lot of this away through unconstrained properties:
hasTranslation/isTranslationOf, hasAdaptation/isAdaptationOf, etc. where the
first is, say, a play modeled in RDA or BIBO or DC Terms and the latter a
film, ballet or score modeled in something else (perhaps oblivious to
FRBR).  Let's say "Romeo and Juliet".  We can assume that something in the
WEM chain is adapted here, but we don't have to know (or care!) what it is,
exactly, to say that Tchaikovsky, Berlioz, Prokofiev, or Zeffirelli's works
are related and how.  That is, we should be able to say that Berlioz's
"Romeo and Juliet", modeled in the Music Ontology is an adaption of (or
inspired by, or whatever is most appropriate) Shakespeare's play, /without/
the W-E-M-I model necessarily needing to be explicitly invoked.  If and when
W-E-M-I and their relationships for these two endeavors are created, that
makes the relationship even richer, but should not be required simply to
link two citations.

This doesn't diminish the value of FRBR in any way, it just lowers the
overhead.


> Hmmm. I wonder what else I can be frustrated about before 8 a.m.? :-)
>

Oh, we can find things...

-Ross.