Re: Non- and Partial-FRBR Metadata

Hello, everyone.

Quick prefatory note: I'm new to this list, and late in entering this
conversation, but I have read it with great interest, and would like to
offer some comments, in response to several different participants.

Ms. Coyle's concern seems to me to underscore the importance of _having_ an
Expression entity rather than the need to work around it:  she specifically
mentions that "The same Expression information may be found in more than one
bibliographic entry", which is precisely a case for normalization.

Later, Ms. Coyle writes: "Empty Expressions add nothing to the usefulness of
our data."  I submit this is not so: firstly, empty expressions are not
empty -- they are identified records linked to at least one other record (a
Work), and usually also to at least one Manifestation; secondly, even if
these relationships are all there is to these expressions today, they _do_
add something to the usefulness of our data -- they add the _possibility_ of
enriching this subset of metadata (i.e. Expression attributes), and the
possibility, sometimes, would yield improved metadata; precluding the
possibility in advance would certainly ensure no fuller Expressions are
available before the next implementation revamping.  This, I believe, was
the gist of Mr. Dunsire's suggestion that implementing Expressions would
be the basis for great potential re-use, as implementations mature and
develop.

I second Mr. Svensson in thinking some sort of "dumb" Expression entity can
be created by the implementation, linked to both the Work and the
Manifestation[s] ("Everything Else" in Ms. Coyle's scenario), anticipating
future uses and navigation/query mechanisms that would make sense of the
Expression level, and until such time, having the current implementation
understand that to go between a Work and [a] Manifestation[s] you traverse
the linked Expression entity/ies.

If, on the other hand, one tries to make do with some sort of MARC-like
flattened set of properties, whether grouped/faceted or not, one has no
_entity_, i.e. no "unique key" in data modeling terms, to "hang"
relationships on.  This does not mean one cannot _generate_ a "unified
record" for a given path of WEM(I) records, for display purposes, but it
seems to me that the power of the conceptual model offered by FRBR
_absolutely depends_ on the separation of the WEMI entities, and so these
"unified" records, even if _stored_ as such (say, for caching results), must
not be the primary bibliographic records in future implementations.

I, too, believe crowdsourcing will have a great role in next-generation
highly FRBRized catalogs and information systems, but the "crowd" in
question need not exclude professionals (librarians etc.).  I expect a lot
of partial and experimental implementations will begin producing quality
hand-FRBRized (or at least hand-verified) records for their own collections
or for subjects of local specialization, at which point the great question
of _federation and interchange_ would become key; this is what this working
group is for, and this is what drew me to join this discussion.  I further
expect free content groups, such as the Wikimedia Foundation, would have a
lot to offer in that respect.

Incidentally, the "lost" Work example for a Work with no Expressions is
indeed interesting, and I can easily think of a real-world use case:
 consider Aristotle's "second" book of the Poetics, the one on comedy; no
versions of it are extant, and some scholars even doubt it was ever written
in the first place.  Nevertheless, there are Works that _discuss_ this lost
Work (e.g. articles speculating on its contents, Eco's _The Name of the
Rose_) and there's even a Work that purports to present a (scholarly, not
fictional) hypothetical version of Aristotle's book (by Richard Janko, ISBN
0715616587) -- this is certainly _not_ an Expression of Aristotle's actual
Work; it is certainly some sort of derived Work, i.e. there would be a
Work<->Work relationship there, and it is _about_ Aristotle's Work (Janko
provides ample commentary and background for his reconstruction), so there
would be an additional, different, Work<->Work relationship between the
two[1].    In a wilder vein, similar arguments could be made about
certainly-fictional works, such as Lovecraft's _Necronomicon_, but let's
leave that out for now.

(sorry about the lengthy post.)

Cheers,

   Asaf Bartov

[1] Somewhat perversely, if we are to express the obvious fact that the lost
work of Aristotle was written in Greek, it seems we _would_ need an
Expression entity after all.  As Mr. Young notes, perhaps we need a "lost"
attribute to adequately express this.

-- 
Asaf Bartov <asaf.bartov@gmail.com>