One can approach this by considering literary/linguistic and web science issues.


Literary/Linguistic -

The need for Work-level resource descriptions is well-justified by literary and linguistic theory, research, and practice.[1] We document key features of the modes of languaging such as speech (inc. its written forms) and sign because they are the means by which concepts are transferred between parties via transient and library-friendly semi-permanent carriers.

Reasoning about this in metadata terms (rather than shifting to more suitable reference frames as needed), consider Work-level descriptions as being able to play a “dark metadata” role WRT Expression-level etc. descriptions.

From a user-focused point of view, Work-level descriptions structure user-facing Expression-level descriptions etc. behind the scenes – and make it possible to discover and access resources in a concept-aware fashion. The strategy is demonstrably effective – but knowledge of it may not interest the casual user. Note that a dark metadata role may change to a visible one if users become more aware ( by encountering multiple copies, editions, expressions of “the same thing,” etc.) of their information-seeking process.


Web Science -

A recent paper on the evolutionary origins of modularity[3] encourages us to conceive of a type of Work-level description that modularizes a number of individual Work descriptions that are asserted to refer to the same intellectual creation. A key finding of this paper is that modularity (chunking) arises in a network environment if a network selection/survival factor is the reduction of connection costs. Since we are talking LLD...

So If Work-level descriptions A, B, ... Z are declared as having a common conceptual basis, a Work-level description that modularizes all of these targets the need to define or compute inter-Work-level descriptions. A single link from a W
(S) description to descriptions A, B, ... Z eliminates the need for additional linkage in the form of situationally-defined onesie, twosie, etc. links among A, B, ... Z Work descriptions.

Naturally, any attributes of or links to a W
(S) modularizing description are also link-economical if those links would otherwise have to been assigned to A, B, ... Z individually.


So Work-level descriptions and modularizations thereof can both play key roles in a BIBFRAME scheme of things – especially with respect to improving resource discovery and link economics.

1 The OED traces literary or musical uses of “work” back to around 1300.


Ron Murray

On 5/16/13 3:24 PM, "Ross Singer" <[log in to unmask]">[log in to unmask]> wrote:


I think you're right, users usually probably want any manifestation in the same language.  Works are useful, of course, to associate the same things in different languages as having been derived from the same original, but I agree that's probably not what people are (in most cases) looking for.

However, I think when it comes to derivative or related works (movies, adaptations, etc.), the work level is critical.

A typical library user is probably far less interested in the fact that the library holds a copy of 'Emma' in Urdu (except maybe the novelty of it) than they are that the library has a BBC miniseries, or "Clueless", or a copy of "Jane Fairfax: The Secret Story of the Second Heroine in Jane Austen's Emma", etc.

This sort of fits into your notion of Work being too abstract, but that's sort of the level of the relationship between two completely different endeavors.  In some ways the relationship between works is far more relevant than the relationship between expressions.


On May 16, 2013, at 1:40 PM, Karen Coyle <[log in to unmask]">[log in to unmask]> wrote:

 I've often wondered about the library use case for a frbr:Work. (It makes more sense to me in other contexts.) Imagine a user going to a library looking for "War and Peace." In most cases, the person wants the book in a specific language, which may or may not be the original language. In a library serving English language speakers presenting the user with "  Война и миръ" probably isn't ideal. Nor would most users want to see all of the different translations, even though that is, under some circumstances, bibliographically relevant.
 It seems to me that the frbr:Expression level is closer to the user view than the Work. The Work, to my mind, is so abstract as to be fine as a topic of discussion ("I'd read War and Peace but it's just too long"), but not a "thing" that people seek to use.
On 5/16/13 9:09 AM, Laura Krier wrote:
 Is a translation really a different "conceptual essence"? I don't think of a translation as a separate work. But being that BIBFRAME doesn't distinguish between Works and Expressions of a Work (which I think a translation would be), it looks like a translation would have to be considered a different Work. What barriers might that introduce to aggregating resources, or discovering resources?


Laura Krier

Metadata Analyst

California Digital Library



On May 11, 2013, at 10:24 PM, Shlomo Sanders <[log in to unmask]">[log in to unmask]> wrote:
Does the BIBFRAME 'work' include different expressions (as in FRBR) - for example different translations?
 Translations may be considered a different "conceptual essence" (albeit a related one) but I haven't been able to find this stated explicitly.    
 Eill it be possible to express links between works - e.g different translations?
 Experience the all-new, singing and dancing interactive Primo brochure
 -----Original Message-----
 From: Bibliographic Framework Transition Initiative Forum [mailto:[log in to unmask] <> ] On Behalf Of J. McRee Elrod
 Sent: Sunday, May 12, 2013 04:44
 To: [log in to unmask]">[log in to unmask]
 Subject: [BIBFRAME] What's an instance?
 I tend to think of a Bibframe instance as an edition, but the Bibframe instance seems to be something different, mainly in envisioning more than one instance per edition, but occasionally more than one edition per instance, using ISBN as the determining factor.
 Often different bindings of copies of an edition have separate ISBNs, but Bibframe says one instance per ISBN.  Since binding is normally not mentioned in description (using "description" in its usual sense, not to mean abstract or summary), how would these instance descriptions differ from each other for the trade, library, deluxe, paperback bindings?  One could have four Bibframe instances for one
 AACR2 or RDA edition.
 Often editions are published simultaneously by two or more publishers, but Bibframe says an instance can only have one publisher.  Sometimes these simultaneous publications have both or more publishers given in the resource.  If both or more publishers appear, surely both or more should be included in one instance description, even if each publisher assigns its own ISBN?  Each ISBN describes the same resource; the only
 difference is who sells it.   An instance description with one
 publisher and one ISBN would not match any existing bibliographic item, each item having more than one.
 Occasionally publishers repeat an ISBN in difference editions.  Are these two dr more editions to be one instance?  Which edition would be described?  How does one handle both in the same collection with only one instance description?  Rarely the same ISBN can appear in editions of different works.
 ISBN is not a safe litmus for determining editions (instances).
 How do yearbooks or multivolume sets, with an ISSN for the serial, an ISBN for the set, but individual ISBNs for the serial and set volumes, fit into this?  (Utlas had 021 for analytical ISBNs of volumes within a serial or set, a feature we still miss.)  While for ebrary, we must create a record for each volume of a mutlivolume set or a serial, because they can have only one 856$url per record, that is not something we would like to do for all.  It would clutter up catalogues.  BTW, can an instance record have multiple PDF URLs?
 If these volumes with their own ISBNs are separate instances, are each instances of a separate work, or are all volumes instances of a single set or serial work?  Instance records for these volumes would seem to have more in common with MARC item records, than AACR2/RDA MARC manifestation records.
 The Bibframe provisions seem to me not to accord with messy bibliographic reality.
   __       __   J. McRee (Mac) Elrod ([log in to unmask]">[log in to unmask])
  {__  |   /     Special Libraries Cataloguing   HTTP://
  ___} |__ \__________________________________________________________