Jonathan,
Thanks for the clarification. (I think these conversations are interesting and enjoyable--mostly!) I think I just conflated the sets approach with the WEMI top-down approach that I feel gets in the way of understanding how the FRBR model can be used (rather than abused) effectively.
I don't think one can get away from the "work" idea though. It--the entanglement of "author" and "title" we call a "work," is so entrenched culturally, we can't escape it. Stephen King's next novel can't be written by someone else. His next novel can't be the same as one that came before it.
One last thing, the traditional library model for cataloging was not a single entity model--of course there was nothing we would recognize as a model or rather as a data model. But what there was mixed many things we'd now distinguish as entities or relationships into a single textual description. It was more a discursive than a data practice.
Well, I see there are many more messages on the topic to read and I have no more time today.
Matthew
-----Original Message-----
From: Next generation catalogs for libraries [mailto:NGC4LIB_at_LISTSERV.ND.EDU] On Behalf Of Jonathan Rochkind
Sent: Wednesday, October 21, 2009 2:13 PM
To: NGC4LIB_at_LISTSERV.ND.EDU
Subject: Re: [NGC4LIB] Tim Berners-Lee on the Semantic Web
I agree with what you wrote Matthew, but disagree that the set
conception requires you to 'start from the top'. In fact, it's just the
opposite, and that's one reason why I think it's useful. You can't talk
about sets unless you first talk about what they are sets _of_. You in
fact need to start from the bottom to talk about sets.
First you can define an item. Then you can define a manifestation as a
set of items; an expression as a set of manifestations (and thus
implicitly as a set of items too); and a work as a set of expressions
(likewise, implicity that means it is also a set of manifestations, or
of items).
I think the set conception in fact encourages and even requires you to
start from the bottom, as you advise. My previous blog post suggesting a
useful way to think about the WEMI entities in terms of sets likewise
starts at the bottom:
http://bibwild.wordpress.com/2007/12/07/frbr-considered-as-set-relationships/
Now, I guess Karen is suggesting that a one-entity model is sufficient
-- we just need the "resource". I am confused as to whether this is
meant to correspond to "item" or to "manifestation". In fact, our
tradition of cataloging HAS been a one-entity model, based on the
manifestation. I think what we have learned from that is that this
isn't sufficient -- we need to organize sets of manifestations because
the user is often not looking for any particular manifestation, and
there is metadata that applies to sets of manifestations that we want to
record authoritatively, not duplicate all over the place on various
manifestation records in possibly inconsistent form. If a user says "I
want a copy of Hamlet", we (and our systems) need to be able to answer
that, not say "Sorry, I can't answer your question unless you tell me
specifically which edition you mean." To do that, our systems need to
understand that there is some set of manifestations that is known as
"Hamlet" and is by "William Shakespeare." Our inherited single entity
model is a serious obstacle to producing data and systems capable of
dealing with this.
If you moved to a single-entity model that was just "item", it would be
even worse, we wouldn't be able to share cataloging data at all, because
we each have different items!
If you move to a single-entity model where it's just "resource" and it
can represent whatever the heck you want it to represent (an item, a set
of items defined any way you like) -- this leads to awfully complex data
that we're going to have trouble doing anything with.
On the other hand, conceptually I can think of a workable model that is
nothing but items, manifestatons, and arbitrary "sets of
manifestations". Sets would be able to have attributes of their own (So
you can label the set of all copies of Hamlet with the title "Hamlet"),
and also sets would be able to have membership in other sets (So you can
take the sets of all manifestations of particular translation of War and
Peace and say it belongs to the War and Peace work set). Each
ManifestationSet would then likely have a 'type' assigned to it from a
controlled vocabulary, so you know what kind of set it is. And guess
what, if you wanted to meet actual needs and be at all compatible with
our legacy data, the first "type" of set you'd need is Work, and the
second most popular would be Expression. So you've just ended up with a
more abstracted form of FRBR. Which might not be a bad idea, but we've
still got to start somewhere.
Jonathan
Beacom, Matthew wrote:
> Karen,
>
> Among other interesting things you said was this:
>
> "All of this being why I would prefer to have a "Resource" entity that has content and carrier, and a lot of relationships like "is expression of" and "is translation of" rather that trying to fit things like "color" into a single box."
>
> And Jonathan said (among other interesting things):
>
> "(And really, I think it all makes a LOT more sense if you consider those 'things', work, expression, and manifestation as sets of items. The FRBR document doesn't make that clear, but neither is it incompatible with that conception)."
>
> I want just now to briefly (that may not be possible for me) disagree with the set interpretation of the WEMI part of FRBR and do that in a way that connects to what I suspect is my own idiosyncratic interpretation of Karen's call for a "Resource" entity. So here goes.
>
> I understand Karen's "Resource" entity to be identical with the FRBR "item." Rather than seeing the WEMI as a hierarchy of sets (each containing the one below), I think it best to flip WEMI so that we start with the item entity. The item entity is the most concrete and specific of the WEMI entities. Even manifestation considered as a set of interchangeable items introduces an new aspect of abstraction or conceptualization that is not part of the thing that is the "item." Manifestation, Expression and Work are best understood as sets of ideas we have about an item and these ideas about an item are, more or less, what I think Karen means by "a lot of relationships." In other words, the item is a real physical thing (even if intangible) and W, E, and M are not. The consequence of this up-side-down view of WEMI is that we can more easily think of W, E, and M concepts that we can define both abstractly and operationally in practical applications of the model rather than spend t!
im!
> e arguing about what the true nature of a "work" is. So we may decide that a manifestation is a particular relationship (and I have no practical trouble with expressing manifestation as either an entity or a relationship--its particles and waves to me) between items. That particular relationship can be defined tightly or loosely, but it is more or less that these items share a production history that makes them more or less interchangeable: one is the same as another in both content and carrier. Although Work and Expression (and I would certainly add "superwork" or "work family" to W and E. as an obvious and useful extension to WEMI)are much more abstract than manifestation, they are still best understood as ideas we have about items. (Though it may be simpler in some cases to think about it as ideas about manifestations or expressions, that is just a shorthand.) Anyway, those ideas may be very usefully labeled relationships. So I think Karen's idea (or perhaps my misread!
in!
> g of Karen's idea) is compatible with the FRBR WEMI entities and their
> relationships. The sets idea Jonathan mentioned is also useful and sometimes practical; its main fault, as I see it, is less the set idea itself that its application from the top down in a hierarchy of WEMI. I have no problem with thinking of a work as that set of all items that share the same author (meant generically to include composer, etc.) and an abstract but recognizable and agreed upon identity of content (which is by-the-way a huge issue and sometimes not many times but sometimes cannot be determined.)
>
> The above may not make half the sense I wish it did, but it is all I can now. See how hard a time I have being brief? My apologies.
>
> Matthew Beacom
>
>
Received on Wed Oct 21 2009 - 17:11:17 EDT