To my mind, in order to move forward in a Linked Data environment, the cataloging process MUST ALWAYS create a Work description for the manifestation described by the primary access point of the description, full stop. (Whether "Work description" in that statement means Work and Expression, or joint Work-Expression, would be determined by the specific coding schema in use.)

That process should be done fairly transparently by the software one uses. (One example: software automatically analyzes every piece of data entered in a cataloging template to determine whether it will be recorded in a Work, Expression (or BibFrame Work), or Manifestation record and does so accordingly, automatically creating & linking Work & Expression records as needed. A different example: A cataloging template contains Manifestation-level fields, plus a way to search for Work and/or Expression records; if one is found, that data is added to the current screen; if none is found, add the relevant template fields in discrete areas of the current screen, with an end result that is functionally identical to what catalogers see now, just with the fields in a different order, and with some fields already filled in when an existing work or expression is found. That's basically how Goodreads currently works.)

Where I see the most major issue regarding the decision of when to create work & expression records is when dealing with manifestations that function as containers.  For example:

* When dealing with pop music albums, should there always be a full WEM hierarchy for each song? ( already does this, for one example.)  Only for certain songs? 

* When dealing with literary collections (short stories, poetry, essays, omnibus editions, etc.), should there always be a WEM hierarchy for each story, poem, essay, etc?  Only for those that have also been published on a standalone basis?

* When dealing with a book + CD kit, a multi-format literacy kit, a collection of 4 different movies on a single DVD (or in a single box), or a Blu-ray + DVD combo pack, should there be a WEM hierarchy for each discrete piece? ( already does this, to some extent.)  Only those that have also been published on a standalone basis?

* When dealing with a serial or multipart monograph, should there always be a WEM hierarchy for each volume/issue?

Describing discrete works vs. describing carriers is a debate that goes back to Lubetzky and beyond, and isn't likely to be fully resolved any time soon, but now that catalog cards aren't an issue, and so much of the basic work of descriptive cataloging has been and continues to be automated, the "extra work" is not as much of a barrier as it was when the 1961 Paris Principles originally pronounced "the book" rather than "the work" as the basic element for description. (Now it's "the manifestation", but since the unstated assumption seems to be that containers are most often treated as a whole, the end result is the same.)

Joel Hahn
Database Manager
CCS, Cooperative Computer Services
Arlington Heights, IL
[log in to unmask]

From: "Ed Jones" <[log in to unmask]>
To: [log in to unmask]
Sent: Tuesday, May 30, 2017 9:51:45 AM
Subject: [PCCLIST] What triggers creation of a Work description in a Linked Data environment? (Question from the PCC SCS/LDAC Task Group on the Work Entity)



[I sent out an earlier version of this post over the weekend, but fortunately it had not appeared on the list as of this writing. It serves as a cautionary tale to me not to do this sort of stuff on an iPad (at least not till my skill level has increased).]


This is the second of the Task Group’s questions, previewed in my presentation at OpCo earlier in the month. Thanks to Yang Wang, Kevin Randall, and Jessica Janecki for their thoughtful replies to the first question. Note that discussion of that question remains open and continues to ramify (see Jean Godby's reply to Yang's reply and Yang's recent elaboration, which will in turn elicit a reply).


The second question has multiple aspects, some anticipated in Kevin's reply to the first:


What triggers creation of a Work description in a Linked Data environment?

  • What conditions must be satisfied?
  • What properties (attributes and relationships) must be included?
  • What services require works?




There should be clear and consistent guidelines to allow programs and catalogers to definitively determine when the creation of a Work entity is warranted.  While this seems as though it should be straightforward and simple, the concept of a Work tends to be loosely defined and the properties that describe a work vary both across and within communities (e.g., OCLC/ creativeWork vs. BIBFRAME Work vs. FRBR Work).  If PCC were to decide that work entities should not be created for singletons or unique resources, what would be the trigger to create a work entity for resources that have multiple expressions or multiple manifestations of a single expression?


Contributing to the ambiguity surrounding the Work concept are vague definitions and a lack of consistent modeling among standards within and across communities. This ambiguity leads to different interpretations of the triggering event for creating a new Work.  In FRBR, for example, multiple FRBR Expressions may be clustered under one FRBR Work, but the guidelines for triggering a new FRBR Work (i.e. what constitutes a new Work rather than a new Expression) are left to the implementation.  Meanwhile in BIBFRAME, FRBR Work and Expression data are combined such that each new FRBR Expression triggers a new BIBFRAME Work.


Such ambiguity leads to more questions than answers about when to create a new work.

  • Do different resource formats require different triggering points for creating a new work entity?  Is the triggering event for creating a new Work the same for serials, music, film, and monographs?  
  • Should a separate work entity be created only when there are multiple Expressions? What are the ramifications from cataloging workflow, data modeling, and machine processing perspectives?  Is there any advantage to creating a Work for every Expression?
  • Should a Work entity be created if there are only multiple Manifestations of a single Expression?  
  • Current cataloging practice expects some descriptive properties of a Work, Expression and Manifestation to be distributed across the WEMI entities.  If Works are created only when there are multiple Expressions or multiple Manifestations, what changes are needed to cataloging rules and modeling assumptions to accommodate ‘Work’ data in a single-entity description model?


At least within the cataloging community, there should to be precise and firm rules regarding the triggering events for creating new Works, even if those triggers vary by format.  Adding clarity to Work description now will not necessarily solve the ambiguities inherent in legacy data but can bring standardization to new cataloging and facilitate future change.




The answer to this question may largely determine the answer to the question of whether Work descriptions can be created by machine, and if so, under what conditions and how autonomously? This question was raised in Yang's latest (May 25) reply. Answering it will have consequences for the extent to which the machine creation of Work descriptions can effect economies in the cataloging process.


Ed Jones,

Chair, PCC SCS/LDAC Task Group on the Work Entity