I completely agree with Karen regarding how FRBR falls short in not
allowing for more relationships between Group 1-2 and Group 3 entities.
FRBRoo fleshes out some of these things, but in a woefully unweildy way,
IMO. Conversely, FRBR in RDF (at http://vocab.org/frbr) consolidates
some classes and properties (e.g. Responsible entity, a superclass of
Person, Family and Corporate body), and to me approaches the kind of
extensibility we need. Unfortunately, it does not include data
properties, which I agree are problematic, as Karen illustrates.
I do maintain that FRBR is the kind of *conceptual* model that, for the
most part, can guide the development of effective data structures.
However, it is far too abstract to be implemented verbatim. This is what
I think RDA is trying to do with attributes like "Title for the work" I
wonder: why is there not an ontology expert on the JSC?? (If I'm wrong
and there is, someone please correct me)
Casey
Karen Coyle wrote:
Extensibility as absolutely key. I know that some people consider XML
to be inherently extensible, but I'm concerned that the conceptual
model presented by FRBR doesn't support extensibility. For example,
the FRBR entity "Place" represents only the place as a subject. If you
want to represent places anywhere else in the record, you are SOL.
Ditto the "Event" entity. The attributes in FRBR have no inherent
structure, so you have, say, Manifestation with a whole page of
attributes that are each defined at the most detailed level. You have
"reduction ratio (microform)" but no "reproduction info" field that
you could extend for another physical format. You have "date of
publication" but no general "date" property that could be extended to
other dates that are needed (in fact, the various date fields have no
relation to each other).
To have an extensible data structure we need to have some foundation
"classes" that we can build on, and nothing in FRBR, RDA, or MARC
gives us that.
kc
Casey A Mullin wrote:
(Attention: lurker emerging)
To me what it comes down to is neither simplicity nor complexity, but
extensibility. In a perfect world, our data models should be capable
of representing very sophisticated and robust relationships at a high
level of granularity, while still accommodating ease of metadata
production and contribution (especially by non-experts and those
outside the library community).
I agree that none of our existing data structures/syntaxes are /a
priori /fundamental or infallible. But what is promising to me about
RDF is its intuitive mode of expression and extensibility (exactly
the kind I advocate above).
Casey
Han, Yan wrote:
Bill and Peter,
Very nice posts. XML, RDF, MARC and DC are all different ways to
present information in a way (of course, XML, RDF, and DC are easier
to read/processed by machine).
However, down the fundamentals, I think that it can go deeper,
basically data structure and algorithms making things works. RDF
(with triples) is a directed graph. Graph is a powerful (the most
powerful?) data structure that you can model everything. However,
some of the graph theory/problems are NP-hard problems. In
fundamental we are talking about Math. So a balance needs to be
made. (between how complex the model is and how easy(or possible) to
get it implemented). As computing power grows, complex data modeling
and data mining are on the horizon.
Yan
-----Original Message-----
From: Code for Libraries [mailto:code4...@listserv.nd.edu] On Behalf
Of Peter Schlumpf
Sent: Thursday, April 09, 2009 10:09 PM
To: CODE4LIB@LISTSERV.ND.EDU
Subject: [CODE4LIB] You got it!!!!! Re: [CODE4LIB] Something
completely different
Bill,
You have hit the nail on the head!!!!! This is EXACTLY what I am
trying to do! It's the underlying stuff that I am trying to get
at. Looking at RDF may yield some good ideas. But I am not
thinking in terms of RDF or XML, triples, or MARC, standards, or any
of that stuff that gets thrown around here. Even the Internet is
not terribly necessary. I am thinking in terms of data structures,
pointers, sparse matrices, relationships between objects and yes,
set theory too -- things like that. The former is pretty much cruft
that lies upon the latter, and it mostly just gets in the way.
Noise, as you put it, Bill!
A big problem here is that Libraryland has a bad habit of getting
itself lost in the details and going off on all kinds of tangents.
As I said before, the biggest prison is between the ears!!!! Throw
out all that junk in there and just start over! When I begin
programming this thing my only tools will be a programming language
(C or Java) a text editor (vi) and my head. But before I really
start that, right now I am writing a paper that explains how this
stuff works at a very low level. It's mostly an effort to get my
thoughts down clearly, but I will share a draft of it with y'all on
here soon.
Peter Schlumpf
-----Original Message-----
From: Bill Dueber <b...@dueber.com>
Sent: Apr 9, 2009 10:37 PM
To: CODE4LIB@LISTSERV.ND.EDU
Subject: Re: [CODE4LIB] Something completely different
On Thu, Apr 9, 2009 at 10:26 AM, Mike Taylor <m...@indexdata.com>
wrote:
I'm not sure what to make of this except to say that Yet Another XML
Bibliographic Format is NOT the answer!
I recognize that you're being flippant, and yet think there's an
important
nugget in here.
When you say it that way, it makes it sound as if folks are
debating the
finer points of OAI-MARC vs MARC-XML -- that it's simply syntactic
sugar
(although I'm certainly one to argue for the importance of
syntactic sugar)
over the top of what we already have.
What's actually being discussed, of course, is the underlying data
model.
E-R pairs primarily analyzed by set theory, triples forming
directed graphs,
whether or not links between data elements can themselves have
attributes --
these are all possible characteristics of the fundamental
underpinning of a
data model to describe the data we're concerned with.
The fact that they all have common XML representations is noise, and
referencing the currently-most-common xml schema for these things
is just
convenient shorthand in a community that understands the exemplars.
The fact
that many in the library community don't understand that syntax is
not the
same as a data model is how we ended up with RDA. (Mike: I don't
know your
stuff, but I seriously doubt you're among that group. I'm talkin' in
general, here.)
Bibliographic data is astoundingly complex, and I believe
wholeheartedly
that modeling it sufficiently is a very, very hard task. But no
matter the
underlying model, we should still insist on starting with the
basics that
computer science folks have been using for decades now: uids (and,
these
days, guids) for the important attributes, separation of data and
display,
definition of sufficient data types and reuse of those types whenever
possible, separation of identity and value, full normalization of
data, zero
ambiguity in the relationship diagram as a fundamental tenet, and a
rigorous
mathematical model to describe how it all fits together.
This is hard stuff. But it's worth doing right.
--
Bill Dueber
Library Systems Programmer
University of Michigan Library
--
Casey A. Mullin
Discovery Metadata Librarian
Metadata Development Unit
Stanford University Libraries
650-736-0849
cmul...@stanford.edu
http://www.caseymullin.com