ALCTS/LITA/RUSA Committee on Machine-Readable Bibliographic Information (MARBI)

Report of the Joint Meeting,
June 17, 2002, Atlanta GA

FRBR and MARC 21

Sally H. McCallum, Network Development & MARC Standards Office, Library of Congress

The Network Development & MARC Standards Office commissioned a study that was executed by Tom Delsey of Thomas J. Delsey Consultancy. The study was designed to map the relationships of elements present in MARC 21 with those identified in the IFLA Functional Requirements for Bibliographic Records (FRBR) model, as well as elements existing within the structure of the Anglo-American Cataloguing Rules. The study also mapped user tasks to data elements.

Historically, MARC has grown by warrant, i.e., by expressed need. Cataloging rules have generally guided the development of MARC, but has never been limited to supporting AACR. The Functional Requirements presents an opportunity to look at the formats as a whole in a conceptual way. This allows us to judge proposed revisions, as well as to identify future directions.

One sub-product of the study was the preparation of a set of examples of FRBR displays. Two possible approaches were identified:

• To display retrieval sets in a way that reflects the FRBR entities, attributes and relationships; or
• To modify databases to reflect the FRBR entities, attributes and relationships.

LC intends to keep the Delsey study up to date as the format are revised and to make it available on the Web. LC is also making available downloadable programs for XML/MARC to FRBR transformations. These programs need to be tested in experiments to determine whether the correct set of data elements has been chosen to identify each entity, and LC encourages experimentation.

As to possible venues for discussion and information exchange on FRBR/MARC issues, the existing MARC electronic discussion list is available. The creation of a new FRBR electronic list should also be considered.

Topics for further discussion include:

• Karen Coyle has raised questions on the MARC discussion list concerning the goals of converting MARC to FRBR? Benefits may to to either end users or to processing staff. Practical experience may be influential and/or informative, e.g., RLG’s long experience clustering records; the WLN database structure; OCLC’s experience with matching and duplicate-detection algorithms; relevant experience at the National Library of Canada; and individual library implementations with various ILS programs.
• How is FRBR related to IFLA’s International Standard Bibliographic Description?
• To what extent will we have cataloging consistency in the future? Do we need a standard level of detail or granularity in description? What are the critical elements that need to be consistent?
• The MARC environment has always been characterized by record independence. What are the implications of FRBR for this construct? European experience is possibly relevant here. A non-independent record environment portends serious issues and implications for record sharing and exchange, which need to be explored.
• MARC has always been broader than AACR or any set of cataloging conventions. An effort has been made to ensure that MARC supports, but is independent of, cataloging codes. MARC would need to support various approaches to FRBRization, and should not be frozen in any one FRBR direction.

We have an enormous investment in MARC standards, existing records, and systems. FRBR is a “young” initiative, but offers interesting possibilities for future enhancement of our ability to present information to users. We need to bring together FRBR, MARC, and cataloging conventions in a winning combination.

Glenn Patton, OCLC

The following issues are related to cataloging rules and practices and the FRBR model:

1. Catalog as a network of relationships: The FRBR model reminds us that a catalog is more than a sequence of catalog records. Cataloging has always been about relationships within and among records. These relationships need to be integrated into how our catalogs operate in the manner suggested by FRBR. Unfortunately, the necessary data isn’t always in our current records. For example, uniform titles have been cited as an example of a traditional mechanism for creating and recording relationships between records. Harry Potter is a relatively simple example of a set of records that might be related through a FRBR structure. On the other hand, Shakespeare’s Hamlet is a very complex example. In many older records, the uniform titles are often lacking, or have been inconsistently applied. Hegna and Murtomaa’s datamining study illustrates how inconsistent past practice has been.
2. Role and function identifiers to indicate relationships: The increased emphasis on relationships in FRBR is not well served by the current practice regarding identification of the roles played by persons and other entities. In AACR and LC practice, identification of roles has been limited. Information that is included in current bibliographic records may only be present in forms not conducive to machine manipulation, [e.g., in natural language statements of responsibility or notes].
3. Form of expression attribute: The FRBR attribute form of expression is closely related to current discussion on AACR rule 0.24 and the “unprincipled” mixture of content and carrier types in the General Material Designations (GMDs). Form of expression, and these related issues, forms an important aspect of supporting the identification of expressions.
4. Work/expression/manifestation mingled in cataloging practice: Current cataloging practice deals with work and manifestation at the same time. We create a description of the manifestation, then make access points for the work(s) and expression(s) that are present. This was a major change in procedure with AACR2. In fact, however, in our shared databases, the first decision a cataloger actually makes is whether the manifestation has already been described, and this involves making decisions about work and expression, even before the manifestation is described. New approaches using the FRBR model may bring us back to the older AACR1 practice which analyzed questions of work and authorship before dealing with description.

Tom Delsey, Consultant

Why map MARC and FRBR?

• This is an exercise in semantics. We need to understand the meaning of the data contained within the format. The formats were designed around the format of the catalog card. Over the past 30 years, MARC has evolved beyond that in its conventions and objectives. MARC now operates in the context of online systems, and faces questions about interfacing with a variety of information providers. Do we still understand the data? It is helpful to look at things from a logical and formal point of view.
• Practically, we need to do intelligent import and export of data from diverse sources. We need to understand and explain our data to communities that do not share our common bibliographic language.
• There is also a practical need to deal with questions surrounding data as we attempt to migrate it to new systems and conceptual structures without loss of functionality.
• A logical analysis of MARC has value regardless of the outcome of the exercise.

Conclusions of the mapping study:

• There is a substantial degree of correspondence between MARC and FRBR model. This is not surprising, since the FRBR model was developed by examining existing structures informed by the MARC model. Specifically, 1200 of 2300 elements mapped to FRBR, although 120 mappings were inexact or anomalous; 200 more could be mapped to FRBR entities, but required adding attributes not defined in FRBR; 150 “wild cards” were too imprecisely defined to map successfully.
• There are many anomalies and uncertainties, particularly at element level (rather than values); many subfields contain multiple data elements (e.g., 533); some definitions are unclear (e.g., access restrictions: to what level do they apply?)
• Finally, about a third of the elements in MARC remain outside the FRBR model. It was necessary to extend FRBR to add new entities and relationships (e.g., “results from,” “used in” relationships).
• AACR mapping: The internal abstract entities in AACR have a major impact on MARC (e.g., class of materials); the heading relationship is to the AACR concept of work, not necessarily to other concepts of work; formal structures are more important in AACR than functional structures (e.g., main entry/1XX).

DISCUSSION:

• [John Attig, Penn State]: To Glenn’s list of discussion issues, we should add the concept of authority records. Many relationships (such as work-to-work relationships) seem to belong in authority records. There is an unresolved conflict between the practice of treating such relationships in a bibliographic record (added entry) or in an authority record (cross-reference); this set of considerations needs to be brought within the model. The work that Tom Delsey will be doing for the IFLA Working Group on Functional Requirements and Numbering for Authority Records (FRANAR) should be very helpful in this regard.


• [John Espley, VTLS]: VTLS has implemented FRBR functionality in Virtua, their ILS product. They used the mappings from the FRBR model and encountered problems deciding when there is a new expression. More specific guidance is needed on this, as well as mapping the concept of edition to FRBR, and on the definition of a serial work.
        [Reply by Tom Delsey]: Work and expression are relative constructs, depending on cataloging conventions, culture, region, etc. Boundaries are defined in “local” conventions like AACR, which relies strongly on the title page. In the case of the concept of edition, a distinction is made between version identification (an attribute of expression) vs. the edition statement (an attribute of manifestation). The edition statement may or may not signal a real change of expression. One cannot reduce this process to a purely mechanical exercise.
        [Comment by John Attig]: In a way this is disturbing because more research is required to determine when a new work or expression is involved.
        [Comment by Glenn Patton]: Concerns about the complexity of the expression entity have been expressed by the Italian IFLA community; in analyzing music works in particular, they find that there may be as many as four distinct levels within the expression entity.


• [Ed Glazier, Research Libraries Group]: How do we cope with legacy records, particularly those that are inadequate for FRBR manipulation?
        [Reply by Delsey]: Many of the significant attributes for identifying the work or expression are present in the record, perhaps not in headings and perhaps not explicitly defined. With FRBR, we are going back to earlier eras and taking useful techniques, such as guide cards from card catalogs and dashed entries from book catalogs. The Italian paper suggests that we develop an order of priority among elements in order to determine the FRBR level. However, with automated systems, it might be possible to offer a user multiple options for organizing the display of expressions.


• [Mitch Turitz, San Francisco State]: For serials, uniform titles serve to distinguish between otherwise identical but unrelated titles. You don’t want to collocate, but to separate. The opposite is true with monographs.
        [Comment by Delsey]: Uniform titles already include elements that refer to works, expressions and manifestations. Thus, serials aren’t the only case where you have uses of uniform titles that may be problematic for FRBR.


• [Paul Weiss, UC San Diego]: Catalogs need to support a variety of functions, beyond information retrieval, e.g., circulation, interlibrary loan, acquisitions, collection development, preservation. We need to consider how the FRBR model would support functionality in all these areas.


• [Jane Johnson, UCLA Film and Television Archive]: It seems that FRBR might be used to identify works, etc., with citations, which might address problems associated with arranging retrieval sets by main entry.
        [Reply by Delsey]: There are deficiencies with the main entry concept in AACR. If we can find other ways of drawing out such relationships than our current practice, we might be able to address these deficiencies. For example, we might be able to do something about the current problem when a change in the order of authors on the title page requires a change of main entry. We should be moving from a prescriptive to a descriptive grammar of relationships, detached from formal structures of main and added entry.
        [Comment by Matthew Beacom, Yale]: The JSC Format Variations Working Group is investigating the use of a “citation” to collocate works and expressions; these citations might ultimately look like either uniform titles or author/title added entries.


• [Matthew Beacom]: We can use FRBR for conceptual thinking, understanding and communication; we can also try to apply them, which requires operational definitions of work and expression in our catalogs, firming up the boundaries between works, etc., through rules and conventions.

Summaries by John Attig, CC:DA Webmaster