Karen Coyle wrote:
What type of entity would be "part" be? I'm thinking that there is no
such entity as "part" but that a work can be a "is part" of another
work. Taking into account that the work is a single entity that may be
related to any number of expression/manifestations it cannot be
"secondary" since that is what it is only in relation to the
manifestation being cataloged. Primary and secondary, therefore, have
to be relationships.
In a sense, a Work is always whole, even if it is part of another
work. If it didn't have "wholeness" it couldn't be a work.
(...)
Yes, that is how I imagine the graph to "grow." But I guess I'm not
sure what the "part" box is in your model -- it appears to be a Work
that has the characteristic of being a part of the aggregate.
Good point.
I think you're right that my "parts" also must by necessity be works (in
the same sense that, say, "The fellowship of the ring" is a work in
itself, which at the same time is placed in a whole/part relationship
with "The lord of the rings"). So in the Nabokov example I don't have
only three works (the two individual works plus the aggregate work,
which has two parts), as I claimed before, but rather five works: the
two individual works, the aggregate work and the two "part works".
I know that having "W1" and "W: Part 1 of Aggregate Work" as two
different boxes next to each other somehow looks redundant, but I still
think this complexity is necessary. Let's look as some more diagrams
which I have just drawn:
http://www.mendeley.com/profiles/heidrun-wiesenmuller/
(among the working papers again, called "Additional diagrams #2", or
directly:
http://tinyurl.com/6o2sh3k
(sorry, it's more than 3 MB; next time I'll compress the graphics more).
The new example illustrates the (fictitious) case of two "Selected
works" editions of Jane Austen’s novels. Both contain the same two works
"Pride and prejudice" and "Sense and sensibility", but one of them
contains the English expressions, whereas the other contains the German
expressions. The two aggregate works were created by
two different persons, completely independently of each other.
Now if you first look at figure 2, which illustrates a straightforward
"work-of-works" approach, you'll notice that starting e.g. with the
"Aggregate Work 2" and going downwards to the work "Sense and
sensibility", you then have no way of knowing which of the two
expressions to take (the English or the German one), and consequently,
there is no way of telling which of the aggregate manifestations shown
at the bottom belongs to this aggregate work.
Compare this to figure 3, which gives the same thing in the alternative
model. I admit that this is much more complicated, but at least it seems
to work: Starting with the "Aggregate Work 2" and going downwards you
first reach the two "part works". These are unambiguously connected to
the expressions which the creator of the "Aggregate Work 2" really used
for his collection (the German versions), and this brings you to the
right manifestation. So I think we need to have this "doubling" of
works, if we want to capture things like that.
Now where are the differences between e.g. "W1: Pride and prejudice",
"Part 1 of Aggr. Work 1" and "Part 2 of Aggr. Work 2"? It is as you
thought: Most of the attributes will be the same, and also some of the
relationships (e.g. the relationship to Jane Austen as the creator).
I think there should be an additional attribute "aggregate"
distinguishing between the individual work (W1) and the "part works".
This would have to be newly introduced to FRBR, and it certainly needs
some further thinking to sort out the details (e.g. do only the "part
works" get this attribute, or also the aggregate work? How can we bring
out the difference between "Part 1 of Aggr. Work 1" and "Part 1 of Aggr.
Work 2", if both get the same attribute "aggregate")?
Another difference - and probably the vital one - between the individual
works and the "part works" is how they are integrated in the network of
FRBR relationships. One difference is, of course, that only the "part
works" have a whole/part relationship with an aggregating work. Another
is that whereas, on principle, all existing expressions of "Pride and
prejudice" are connected with the box for the individual work, only the
expression (or expressions) really used for the aggregate work is/are
connected to the "part work". A case where the "part work" boxes would
be connected with more than one expression would be a collection of
essays, which is republished in a revised version (including revised
versions of the essays). Then each part work for an essay would be
connected with two expressions.
One thing I don't like about the diagrams of the alternative model is
that there is, as yet, no direct line between the box for an individual
work and the corresponding "part work" boxes. I feel there should be a
relationship there of some kinde but am uncertain what it might be.
There doesn't seem to be anything adequate in FRBR yet; perhaps this
would also call for an extension of FRBR.
By the way, it was correctly pointed out to me off list that the model
would have to be recursive, e.g. allowing for an aggregate work (e.g. a
collection) being part of another aggregate work (e.g. a monographic
series). Now, that'll be some _really_ complicated diagrams, but I
believe it wozld be possible to model this as well.
I also note now that your Fig. 3 has an expression that realizes more
than one work, which I believe is problematic. It definitely violates
the current FRBR model, but then you are advocating for change in that
model.
Yes, I'm aware of the fact, and that would be the most important change
necessary to FRBR in order to implement the alternative model.
I'm not sure what you mean with "title search" here. Do you perhaps
mean a title search on manifestation level? That's not what I have in
mind. I rather imagine a system like OCLC's FictionFinder (by the
way: will that ever go online again?), which at the first step
presents not manifestations, but only works.
But I believe it searches on all titles. Otherwise, one would have to
know the original language title in order to retrieve the work.
Unfortunately Fiction Finder doesn't seem to be running at the moment
so I can't check that. The other option is that all manifestation
titles would need to be alternate titles in the work.
It hasn't been running for a long time. But I think how it worked was to
cluster all the manifestations to works first. You are right, probably
what they did was then store all titles of manifestations in the work
record, so for retrieval it doesn't matter whether you search with the
original title or, e.g. the title of a manifestation in another language.
However, I don't think we can design for a single system structure.
Surely some systems will provide a full keyword access on any entities.
True, in this case the system would be a bit more cumbersome as you'd
probably have an additional step to take. But this is a matter of
implementation, not a question of principle. I think we should aim at
providing sound data in sound data structures. We can, of course, give
some hints as to possible ways of using the data in retrieval systems,
but I'm not sure whether we can do more than this.
No, I actually think we're getting very close. It would be useful to
have examples, so if you can mock up examples of your ideas I think
that would help. Then we can refer to specifics. What I really want is
a real time white board for drawing diagrams... this kind of thing is
very hard to do in email.
Yes, that would be good to have. For my diagrams I use Cmap Tools:
http://cmap.ihmc.us/download/
It seems that this can also be used collaboratively, but I haven't tried
that out.
(And I greatly appreciate your excellent command of English, as there
would be no communication at all without it.)
Thanks (blushing again). I've got an M.A. in English which comes in
quite handy.
Heidrun
--
---------------------
Prof. Heidrun Wiesenmueller M.A.
Stuttgart Media University
Faculty of Information and Communication
Wolframstrasse 32, 70191 Stuttgart, Germany
www.hdm-stuttgart.de/bi