On 6/5/12 10:51 PM, Heidrun Wiesenmüller wrote:
Thomas said:
The Find user task needs to be satisfied. In card catalog
conventions, the main entry heading collocates related works. Using
some sort of (standardized) method for the value of the "Work
manifested" means that other works can specify something that will
link back to the work in this mix of data. The objective in RDA 17.2
is:
"Find>>all<< resources that embody a particular work or expression"
-- this implies some convention is needed beyond a set of loosely
related elements like Creator and Title proper.
I very much doubt that having a name-title string in the data is an
adequate tool to satisfy this user task, at least unless there are
additional mechanisms in place.
I agree. I also wonder about the "find all" and concur with your
conclusion below that most people looking for a work want to find A copy
of that work, not ALL copies of the work. Note that Cutter's famed user
tasks were worded like:
"1. To enable a person to find a book of which either
(A) the author
(B) the title
(C) the subject ... is known"
Note that he says "a book" not "all books," although this could just be
a turn of phrase on his part. Yet, headings in his word serve not to
retrieve but to "collocate" -- that is to put the entries near each
other in location. There is no expectation that the user will (or must)
view every card before making a selection, and there is no concept of
"find all," as far as I can see. That concept seems to have come about
with retrieval, and retrieval depends on concepts introduced by the
computer.
3. Work clustering algorithms:
In a first step, only _one_ manifestation of the work in question has
to be found (it doesn't matter which). In a second step, all
manifestations belonging to the same work are retrieved by making use
of a work clustering algorithm.
I have recently been thinking about Work as a cluster rather than a
*thing*. It may not be analogous, but my mental model is that of VIAF,
where authority records are clustered such that the cluster represents a
named person or corporate body, etc., in spite of minor differences in
cataloging practices. The clustering does not change the content of the
clustered records, which to me is an important elements.
I haven't figured out where Expression fits into this model, but it
makes sense to me that a Work is a cluster or set of all of its
Expressions/Manifestations; and that a Work does not exist unless there
is at least one Expression/Manifestation/(Item?).
I did a blog post with the beginnings of this thought in 2009:
http://kcoyle.blogspot.com/2009/08/what-is-frbr-work.html
To me, this bottom-up approach makes more sense than a top-down one
because we only really encounter Items in real life -- the rest is
abstract. With you, I wonder if WEM isn't something that we interpret
from bibliographic data, assuming that the data contains appropriate
elements. Given a large corpus of bibliographic data, I would think that
such clustering could be nearly as accurate as would be created by human
catalogers, and would be much more efficient in terms of time.
One advantage of clustering, in my view, is that bibliographic items can
be clustered based on different criteria if desired. Thus communities
that have a different view of Work or Expression from the *standard* RDA
view can see the Work that meets their needs without having to
re-catalog or to translate/crosswalk data. This seems more flexible than
having just one definition of Work/Expression/Manifestation -- instead,
it allows interoperability between different such views.
kc
Method 3 may be especially interesting. In German catalogs, which
already have links to authority records for persons and corporate
bodies (which means that variant names can be used for searching), it
should be easy to retrieve at least _one_ manifestation of a certain
work, even if a variant title for the work is used (provided the
catalog is of a certain size and also includes a fair sample of
material in different languages). Also, there are already
implementations of work clustering algorithms which work reasonably
well. For instance, in some Primo catalogs, other manifestations of
the same work are automatically retrieved and can be displayed by
clicking on a button. Of course, we also know the "view all editions
and formats" link from WorldCat.
I believe that ideas like this are one of the reasons for the German
decision to implement RDA in scenario 2 instead of aiming at scenario
1. It is felt that FRBRization can be achieved by technical measures
and be shown on a surface level (only "virtually", as it were),
without having to change too much in the underlying data structures
itself.
Having to create an authority record for _every_ work, as it would be
necessary for method 1 (and perhaps also for method 2, as the work
information obviously must come from somewhere), seems to be a
daunting task. But now I come to think about it, I wonder: Wouldn't it
be possible to generate work authority records automatically? Based on
work clustering, we could e.g. collect all variant titles for a work
from the various manifestations. Maybe this is a direction worth
looking at.
I'm sure there are more methods than the three I've mentioned. And of
course, linked data, could also play a part (one possibility of
handling variant personal names could be a tool which makes use of
VIAF in RDF, for example).
Heidrun
--
Karen Coyle
kco...@kcoyle.net http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet
