[open-bibliography] FRBR examples

[John Mark Ockerbloom]

This has been a fascinating discussion, one that I've only now had the
chance to review in detail.  I agree with Tim and a number of the other
posters that what ultimately matters is how well the models we adopt
apply for practical problems and tools.  But there are a variety of
problems and tools that are out there, and the more they can share common
information, the better.

One problem that I haven't heard brought up explicitly in this thread,
but that's of particular relevance to OKFN groups like the Public Domain
subgroup, is the issue of rights determination.  The copyrights
that apply to a book are usually at neither the highest or lowest
level of abstraction for a work; in the US system, copyrights usually
apply to a particular expression, but sometimes rights also apply
more concretely (as with the "published editions" shorter-term
copyright that some countries have), or more abstractly (as with the
limited copyright protection that's given in many countries to
distinctive original characters and settings, apart from any particular text
that features them.)

With regard to abstract entity models, I've found a fairly abstract one useful`
for various views on The Online Books Page.  At the risk of reinventing
a wheel that's already been discussed to death by librarians, I'll
describe it briefly here.  It basically boils down to these two
types of statements:

   <Entity1> SPECIALIZES <Entity2>, in <Relator> way
   <Entity1> RELATES-TO  <Entity2>, in <Relator> way

I use this for subject browsing, for example, where SPECIALIZES refers
to BT/NT relations in LCSH, and RELATES-TO refers to RT relations in
LCSH.  The <Relator> information, which is based on a controlled vocabulary,
is useful for selecting and ordering relationships; for example, in my
subject browsing views, I distinguish between different ways that narrower
terms or related terms are declared or inferred, in order to sort and present
them in an order that makes sense to the reader.  See for instance:

http://onlinebooks.library.upenn.edu/webbin/book/browse?type=lcsubc&key=Ontology

where the direct subdivision-based LCSH specializations of "Ontology" are shown
before the specializations that were declared as NTs.

In a bibliographic display, I'm considering using SPECIALIZES relationships to
consolidate results, so that if Entity1 and Entity2 are both part of a
set of search results, and Entity1 SPECIALIZES Entity2 in an appropriate way,
I might only display Entity2 in the results.  (If I implemented this with the
URL above, you'd see Plato's Sophist appear only once under "Ontology", probably
higher up the hit-list.)   This gives the user a more concise results list, with
the option of drilling into the entities "underneath" Entity2 if the user wants
to get more specific.

This sort of model is compatible with FRBR's WEMI stack (Manifestation1 
SPECIALIZES Expression1, which SPECIALIZES Work1), but it doesn't *mandate* it.
[The distinction between SPECIALIZES and RELATES-TO is that SPECIALIZES
implies (possibly overridable) inheritance of applicable properties
of the specialized entity, and RELATES-TO doesn't.]

This model is fuzzier in some ways that folks might object to,
but I think that any system that involves highly participatory shared
cataloging is going to have to accommodate a fair bit of fuzziness.
Moreover, it provides an incremental migration path from today's
world of MARC records that contain everything about particular
manifestations: you move some of the shared information into higher-level
entity types when it makes sense, and leave it inline in the bib-record
when it doesn't.

I do wonder whether something like this, along with widely used
conventions on relators and essential concepts, might become the de-facto
framework for shared open cataloging.

John