RDA Models for Relationship Data

8 August 2016

RDA Models for Relationship Data


Submitted by: Gordon Dunsire, Chair, RSC Relationship Designators Working Group

Print Friendly, PDF & Email

Tagged , , , . Bookmark the permalink.

37 Responses to RDA Models for Relationship Data

  1. Tina Shrader says:

    General comment: I enthusiastically support this paper’s move toward cross-entity relationships. Being able to have relationships between, for example, the manifestation of one work and the expression of another work has the potential for much greater flexibility to deal with the kinds of complex situations one finds when dealing with works in series and/or serials.

    • Robert L. Maxwell says:

      I totally agree. Any entity may potentially have some sort of a relationship with any other entity, and therefore no relationship should be forbidden. I’m not sure the extremely complex method of dealing with this in this paper, with an attempt to create a qualified designator for each possibility, is helpful, though.

  2. Tina Shrader says:

    Recommendation 1: I support the establishment of a generalized ‘related entity’ structure to support more specific relationships within RDA.

    • Tina Shrader says:

      Passing along comments from my colleagues at NLM:

      Recommendation 1: I’m OK with this and strongly support 1a and would like the community to encourage more use of a general “there is an association between these two res” but not agonize as to what the specific relationship is. Kind of like an Amazon recommendation—“If you are interested in this this, you are probably also going to be interested in that, without a long explanation of why. Sometimes knowing the category of relationship is important, but other times not. I’m still not sure that as written, this recommendation allows full cross-entity relationships. Sounds like they are still going to be restricted to ones specifically defined, like Reproduction. This seems short-sighted.

  3. Tina Shrader says:

    Recommendation 2: I strongly support adding subject relationships for Agents (Person/Family/Corporate Body)

    • Tina Shrader says:

      Passing along comments from my colleagues at NLM:

      I definitely support recommendation 2 that persons, families, and corporate, bodies should have subject relationships to works. However I do not understand the wording they chose in the proposed revision of appendix M. Again these seem backwards to me. A person is described in a work so why is person in parenthesis?
      described in (person) A work that describes a described person. Reciprocal relationship: description of (person)
      description of (person) A person described by a describing work. Reciprocal relationship: described in (person)

      If the relationship is between a work and a person how can the reciprocal relationships be only to the person?

      • Kathy Glennan says:

        Yes, this would make more linguistic sense as:

        described in (work)
        description of (person)

        but then “described in (work)” would be repeated with different definitions/scopes — one for person, one for corporate body, etc.

        That would likely lead to the dual qualifier approach presented as Change 3.

  4. Tina Shrader says:

    Recommendation 3: I think defining reproduction relationships as being cross-entity between items and manifestations makes sense.

  5. Tina Shrader says:

    Recommendation 4: I agree that reciprocal designators are useful, though some of the specific terminology used in the designators and definitions seems awkward

  6. Tina Shrader says:

    Recommendation 5: I think I agree that organizing relationship designators first by the entity being described and then by the related entity makes sense, though again, I’m confused by some of the terminology and structure in the definitions of specific designators.

    • Tina Shrader says:

      Passing along comments from my colleagues at NLM:

      I agree that we will need several ways of representing elements and designators. Re the issue of creation of new designators, I would like to see this slow to a trickle. I agree with the statement on p. 15 that “creating … designators for every variation found it manifestations is not sustainable and impairs utility for the cataloger and end-user.” Would it be so horrible if “editorial director” and “compiler” were subsumed under the relationship “editor”? You would still have the transcribed SOR to provide the wording on the piece. I find the wording used as examples for relationships and reciprocals to be awkward at best if not incomprehensible. (Perhaps this is actually a comment on Recommendation 6)

  7. Tina Shrader says:

    Recommendation 6: I agree that the organization of relationship designators in the RDA Toolkit should be carefully thought out, as should their labels. I think I prefer the alternate labels in the second iteration of Appendix M.

  8. Kathy Glennan says:

    Change 1

    Note that in 29.1.3, the end of each definition now ends with “… being described.” The current wording uses “… being identified.”

    I assume this is a reasonable rewording in light of the broader changes undertaken here. Does anyone feel differently?

  9. Kathy Glennan says:

    Change 2

    Although the paper asserts on p. 9 that the qualified labels are confusing for end users (although they’re not designed as display labels), there really isn’t any other end-user solution proposed here — I think. The results in this paper may make sense from a machine-manipulation and predictability standpoint, but “described in (person)”, for example, just doesn’t make any sense out of context.

    I’m not sure there’s a solution here, but going down this path and trusting that systems designers (or someone else) will do something more helpful to the end user is a very trusting perspective indeed.

    • Robert L. Maxwell says:

      Completely agree. And as I noted below under “The Qualified Labels”, the machine can perfectly well figure out, without a qualifier, what’s at either end of the relationship– E.g.: [Person description]-[Relationship]-[Corporate body description] — it should be perfectly evident to the machine that this is a relationship between a person and a corporate body without the need of any qualifier, parenthetical or otherwise, in the label for the relationship.

    • Kathy Glennan says:

      From John Hostage (posted to [rules] list):

      Change 2 on page 19 seems to have some typos. I think in each pair of designators, the second one should have the qualifier (work), e.g. description of (work). The document states on page 9:
      Labels for designators for similar relationships between different pairs of entity are distinguished by adding the name of the entity (e.g., “(Expression)”) to which the designator applies; that is, the entity being described given as the domain of the property.

      The reciprocal relationships all use (person), but I think they should be (family) or (corporate body) or (work), as the case may be.

    • Amanda Ros says:

      I agree. As for trusting that someone else will do something more helpful, I’m not that trusting.

  10. Kathy Glennan says:

    Change 3

    For this new set of relationships, I prefer the 2nd option, as the new section J.5.6.

    I do not like either of the solutions presented for using two sets of qualifiers to disambiguate these terms. I have a hard time understanding what these phrases mean, even with the background presented in the paper. For example:

    reproduced as (manifestation) (item)
    related manifestation reproduction of item

    Question: in Appendix 3, why is the term “related” introduced? These are all relationship designators after all; none of the existing terms use “related”

    • Kathy Glennan says:

      Definition questions:

      * Is “of another manifestation” really needed in the new Equivalent Item Relationship definitions? I think there’s only one way a manifestation can reproduce an item — and that’s if it is from a different manifestation. Of course, the inverse is not true.

      * In “reproduction of (item) (manifestation)” the indefinite article “a” should be added before “manifestation” in the definition, creating the following: “…as the basis for a manifestation…”

      * I’m struck by the use of “reissue” in the definition of “reprinted as…”, but this is from existing RDA wording.

    • Kathy Glennan says:

      OK, well the broadest terms now in the Glossary do use “related” (like “related work”), so perhaps that is where this comes from. However, it’s hard to see how this is value added, or even really informative.

    • Robert L. Maxwell says:


      [manifestation description]-“Reproduction of”-[Item description] should be all the machine needs to understand that this is a relationship between a manifestation and an item. I see no need of either incomprehensible double qualifiers or incomprehensible long phrases.

  11. Kathy Glennan says:

    Recommendation 1

    Recommendation 2

    Recommendation 3
    Agree, but as noted above, the proposed terms make little sense.

    Recommendation 4
    Agree, but I should note that these look very strange to my AACR2-trained eyes:

    programmer: A person, family, or corporate body responsible for creating a computer program. Reciprocal: programmer of
    programmer of: A work that involves a creation responsibility for a computer program. Reciprocal: programmer

    surveyor: A person, family, or corporate body contributing to an expression of a cartographic work by providing measurements or dimensional relationships for the geographic area represented. Reciprocal: surveyor of
    surveyor of: An expression of a cartographic work that incorporates a contribution of providing measurements or dimensional relationships for the geographic area represented. Reciprocal: surveyor

    Recommendation 5
    Further consideration of how to present this information (and possibly in different forms) is a good idea. Personally, I don’t find the single alphabetic list in the revised Glossary to be helpful as a cataloger.

    Recommendation 6
    I welcome further review of these labels.

    • Robert L. Maxwell says:

      Recommendation 1

      Agree, as long as this doesn’t imply getting rid of the specific designators.

      Recommendation 2


      Recommendation 3

      Agree (and agree with Kathy)

      Recommendation 4

      Agree, but they must be understandable (i.e. ditch the qualifiers, which make no sense)

      Recommendation 5

      I agree with Kathy; I don’t like the idea of a single list but see no harm in investigating other possible arrangements.

      Recommendation 6


  12. Kathy Glennan says:

    Requirements for new relationship designator proposals

    The paper lays out (p. 15) a list of 7 things that will be required for new relationship designator proposals. This uses terminology familiar to those in the RDA Technical WG, but likely unfamiliar to most catalogers.

    What will the process be to get proposals into this form? Is it expected that catalogers will understand this sufficiently to prepare a proposal, or will it need to be vetted by someone who can “translate” this into the more technical approach outlined here?

    • Robert L. Maxwell says:

      As noted below, I believe the proposed list of requirements sets an impossibly high hurdle for all but the most sophisticated of catalogers. If this is implemented, people will be unlikely to proposed new designators, which would be a shame.

  13. Kathy Glennan says:

    Does this paper move us toward ending the designator proposal moratorium?

    I’m concerned that while this paper lays out some very important considerations for moving RDA forward and dealing with relationship designators in a more coherent and formalized way, I don’t see an immediate path forward for lifting the relationship designator proposal moratorium that took effect after last year’s JSC meeting. I think this is a big problem, both for catalogers and users.

  14. Robert L. Maxwell says:

    The Qualified Labels

    On p. 9 the document notes that RDA Toolkit users find the qualified labels using parentheses confusing. This is the understatement of the year. Worse, not just RDA Toolkit users, but more importantly database users find them even more confusing, and yes, it will only increase if there are double qualifiers. We can say that the qualified labels are not intended for end-user consumption all we like—they will inevitably be displaying to end-users. I know of no current system that does not display them in their entirety. So “confusing labels” is a major problem. The alternative labels in appendix 3 are, if anything, even more confusing. I think your average cataloger (including me), to say nothing of 99% of our end-users, will be extremely puzzled by something like “related manifestation digital transfer of item” or “related item facsimile of manifestation.”

    The recent discussion on the RDA list of the three new designators that suddenly appeared in Appendix K with no discussion at all, authored by the same person as the labels in this document, such as “founded corporate body of corporate body”, show to me that somebody else, perhaps, should be entrusted with devising the designators.

    My question is: why do we need to qualify these at all, whether parenthetically or with a phrase? Since the recommendation is to allow relationships to be recorded from any entity to any other entity (and I fully support this), there is no question that any particular entity to entity relationship that any cataloger might attempt to record will be “wrong” in the sense that the relationship is not permitted;

    And database systems are not so stupid that they need a qualifier on the relationship designator to figure out what kind of relationship it is: the very fact that a work description is linked to, e.g., a person description via a recorded relationship tells the system that this is a relationship between a person and a work—so I don’t see why the label needs to be qualified as in “related work (person)” or “related work of person”. The system can figure out that it’s a person to work relationship simply by the fact that an instance of person and an instance of work are linked. The label is to help human beings figure out what the relationship is, and so having confusing labels is self-defeating. I should think “related work” and conversely “related person” would work without all the qualification.

    The document on page 17 under “Impact of revisions” says that the revisions will “serve the user to fully relate RDA entities”—well, yes, but only if the designators are not so opaque and confusing as to make them unuseful. I do wonder what a user would make of “related expression act” or “related expression acting”—labels like that do not serve the user. I have no idea what those mean. I’ve read the document myself fairly carefully and I still don’t understand what “reproduced as (manifestation) (item)” is supposed to convey. And I remind us all that, while in the future perhaps all relationship designators will include a scope note and a definition (at least I certainly hope so), our users will not be consulting that definition or scope note. All they have to go on is the words in the label, and so we have to get the labels right. They must be clear and not confusing. The labels proposed here aren’t really either.

    • Tina Shrader says:

      Thanks to Robert for stepping back and questioning the basic premise that the qualifiers are needed. I agree that if all cross-entity relationships are now valid, the qualifiers may be redundant and unnecessary. I didn’t read the proposal as being quite as permissive about allowing any entity to be related to any other entity as Robert did, but I would also support that as a goal, for all of the reasons he suggests here.

    • Lori Robare says:

      From Kevin Randall:
      I have not seen any explanation, convincing or otherwise, for the requirement of domain and range qualifiers in relationship designators. As Robert Maxwell has repeatedly pointed out, the domain and range are evident from the related entities themselves; if the entity described (subject in the RDF triple) is a work and the related entity (object in the RDF triple) is a person, any software making use of the relationship designator should be able to determine that the subject is a work and the object is a person.

      The only explanation I can imagine is that we need to have the domain and range specified when using the designators in non-RDF scenarios. The most obvious of these would be the MARC record. While the MARC authority record does distinguish between person, family, and corporate body (as these are currently defined in RDA), it does not distinguish between work, expression, manifestation, and item. The MARC bibliographic record is an uneasy conglomeration of work, expression, manifestation—and sometimes item—data, which fulfilled the FRBR user tasks reasonably well in the less sophisticated discovery technologies in use up until the late 20th century. But the MARC formats are not up to the task of identifying and storing the RDA data elements sufficiently to take us further along in meeting the FRBR user tasks in modern and yet-to-be-developed discovery systems.

      I consider the specification of domain and range entities in the RD label to be essentially a replacement for creating specific MARC tags or codes to identify the domain and range for the related entities in a MARC record. In an RDF triple, there would be no need to make those identifications, because the domain and range would be inherent to the actual entities related. Do we actually need to make RD labels so specific, solely to accomodate the MARC format? And if we do, will that actually improve our ability to manipulate the data? Are we expecting MARC records to be able to function as well as linked data? If so, that makes me wonder about the need to moved into linked data. (If anyone is wondering, I think that the answer to the last question here is “no”—and I fully support the movement toward linked data…)

      If there is any explanation for domain and range qualifiers in RDs that I have overlooked, I would appreciate someone pointing me that way. I’m all ears (eyes).

  15. Robert L. Maxwell says:

    Requirements for new designators

    On p. 14-15 is a list of requirements for all new designators. This seems like an extremely high hurdle to expect catalogers to leap over when they see the need for a new designator, and would in my opinion discourage proposals at the least, or perhaps stop the proposal process entirely. It would certainly slow it down (and given the moratorium we’ve already slowed it down enough). Most catalogers are highly intelligent professionals and know when a new designator is needed, and have pretty good suggestions about what that designator should be; but they probably don’t know a “range” or “domain” from a hole in the ground. The proposal process needs to include what is reasonably necessary (I agree with including a definition of the relationship and scope, for example), but shouldn’t rise above what is very likely impossible for most catalogers. We want to encourage, not discourage, proposals.

  16. Robert L. Maxwell says:

    What is the goal, anyway?

    When the 2015 ALA Appendix K proposal was not approved and the moratorium was placed on consideration of new relationship designators, my understanding of the reason was that a theoretical foundation was needed to help us judge whether newly proposed designators were appropriate, whether they “fit in” correctly into the overal scheme. That is fine. What I’m wondering, though, is how this paper helps move us toward that goal? I’m intimately familiar with the ALA proposal, which contained relationship designators that were considered by all the members of the ALA task group as well as others in ALA to be needed designators for cataloging being done right now. So whenever the moratorium is lifted they are going to be presented again—because they are needed. I don’t really see anything in this paper that’s going to help me and the ALA task group reevaluate what we proposed. As a practical matter the moratorium can’t go on forever or even much longer. What we need are principles that will help us evaluate the existing relationship designators and proposed new ones. I thought that would be one of the main outcomes of the Relationship Designators Working Group’s work. I don’t see that in this paper.

Leave a Reply