FRBR Workshop

    Program and proceedings


Prepared by Barbara Tillett and Patrick Le Bœuf

Session A: Aggregates in FRBR

Ed O'Neill: "Relational Models for Aggregates"

"Clear guidelines for identifying components, works, and aggregates are a key to developing the full potential of FRBR." O'Neill described works as elemental, simple, or complex. He proposed excluding recursive relationships.

Ed O'Neill suggested that the 3 following definitions be included in reviewed FRBR:

  • Work: The smallest autonomous intellectual or artistic entity.
  • Aggregate: An entity comprised of multiple works. Three models of aggregates were presented: 1. single work, 2. work of works, and 3. (his preference) manifestation of works.
  • Component: A subunit or component part of a work. This entity is recursive: any component can in turn consist of smaller components.

Eventually, Ed O'Neill had 4 recommendations for the FRBR Review Group:

  • Recognize the universality of works ("if ever a work, always a work"),
  • Define works as the smallest autonomous entry,
  • Exclude recursive relationships for works,
  • Treat aggregates as manifestations.

O'Neill used the examples of Humphry Clinker (48 expressions, 274 manifestations (with 190 in single work manifestations and 78 in aggregates) and Simon & Garfunkel's "Bridge over troubled waters" in multiple manifestations and expressions over time.

During discussion several people disagreed with this view of works and aggregates, pointing out that how one chooses to implement the model is separate from the conceptual model itself, and that conceptually work is recursive and is contextual (can't say if ever a work, always a work, as it will vary depending on the user view and application).

Maja Žumer: "Modeling Augmentations"

Maja Žumer suggested that a Manifestation should be deemed to embody as many Expressions as there are distinct parts corresponding to distinct types of Works (text, illustration, music), rather than just one "main" Work (traditionally, the textual work) for which a new "augmented" Expression is defined. In some cases, however, a new, composite Work (consisting of textual, graphic and musical aspects) can be recognized, if needed. The important thing is that all works (and expressions) deemed to be important for our users are recorded.

She pointed out that augmentations may not be the best term, because there is not always a primary or main work that is being augmented. She used the example of a children's song book, "Philadelphia chickens", that includes lyrics, drawings, music, and a CD sound recording.

Judith A. Kuhagen: "Modeling Continuing Resources in FRBR (and more...)"

Judith Kuhagen suggested that we consider that all segments of a serial (represented by one or multiple bibliographic records due to different methods of accommodating seriality) together equal one serial work; that we recognize that reproductions may not be issued/released with the same mode of issuance as the "original" and that they should be considered as the same to satisfy the user's need to find and identify (they are just a surrogate for the original).

Other resources have seriality: finite integrating resources and multipart monographs are also likely to change over time.

Judith Kuhagen suggested some adjustments to the FRBR model:

  • Acknowledging seriality of resources (noted in "Areas for further study") (the attributes of a manifestation are now only defined for concurrent manifestations: Different title, place of publication, publisher, series, physical medium, etc.; we need concept of consecutive manifestations of same work/expression: Change in title, place of publication, publisher, series, physical medium, etc.);
  • Reconsidering some attributes (the CONSER Task Force on FRBR and Continuing Resources has already suggested that "Expected regularity of issue (serial)" and "Expected frequency of issue (serial)" be moved from the Expression level to the Manifestation level; in addition, "Edition/issue designation", now at the Manifestation level, should also be at the Expression level for some serials);
  • Another mode of issuance: streaming resources (e.g., news feeds that change in real time), should be recognized.

She reported on a non-bibliographic model of using work authority records to collect the elements needed to find works that are continuing resources and to link such authority records to holdings records.

Carol van Nuys and Ketil Albertsen: "Modelling web resources"

The speakers described the Paradigma Project in Norway and their use of FRBR and suggestions for a digital application. They suggested the concept of "box classes" could be added to the model, for Web resources. This would address the finding that is was not sufficient to have implicit aggregates — it was too costly and they needed to automate the attributes. Their creation of synthetic "boxes" to hold Web objects enabled them to put objects in several boxes and clearly identify each object. They described 8 types of relationships and the box classes and box class properties, modeling the "real world" as their solution for their Web archive. Unfortunately the project ended due to lack of funds. Their concept of work met a functional need for identifying intellectual property.

Considerable discussion followed regarding the ability to describe components and the expense of analytics.

Session B: Review of FRBR and Relationships in FRBR

Allyson Carlyle: "FRBR: challenges for implementation in AACR2, with some attention to non-book materials"

Allyson Carlyle identified the following 3 challenges:

  • fully incorporating the "work" entity into cataloging rules at last
  • accommodating "expression" in rules
  • addressing the weaknesses in the rules.

The boundaries set by rules for when you have a new work or when you have a new expression need to be re-examined. Should we consider a movie and a book as embodying the same work? Perhaps display clarity is more important. Carlyle cited Martha Yee's 1982 article about a general rule for entry of works of mixed responsibility and also cited Patrick Le Bœuf's article about bona fide and fiat works. In AACR2 rules, catalogers do not now record the data we need to identify expressions, and by practice for example we do not record translators or illustrators for expressions. We could never write a rule for every case, but we need to give reasons for rules and provide context. She suggested the need for research to discover the user's differences in perception.

Barbara B. Tillett: "Relationships in FRBR"

Barbara Tillett suggested that one additional "User Task" be defined: "to relate"; besides, other possible user tasks might include: "to attribute royalties to", and "to preserve".

Some relationships are only implicit in the FRBR model, and could be made explicit:

  • if a Work is about a given topic, all Expressions, Manifestations and Items pertaining to that Work inherit that subject relationship;
  • if a Work is realized by 2 or more given Expressions, those Expressions have an implicit sibling relationship;
  • if an Expression is embodied in 2 or more given Manifestations, those Manifestations have an implicit sibling relationship;
  • there is an implicit sibling relationship among all of the copies (Items) related to the same Manifestation;
  • content relationships defined at the Work level are inherited by all other entities from Group 1;
  • whole/part relationships imply relationships among parts (parts can have a sequential relationship or a companion relationship, either dependent or independent).

Other relationships, which are not even mentioned in the FRBR Final Report, could be defined as well: for instance the "shares an attribute in common with" relationship (all Manifestations embodying at least one textual Expression in a given language are interrelated through that relationship).

Maja Žumer: "Some outcomes of the CRM/FRBR harmonization: the definition of manifestation and a review of attributes"

Žumer reported on the CRM, object oriented model of the museum community (see also Patrick Le Bœuf's talk in Session E). The museum analysis indicated the attributes typically reflected in bibliographic records are not an exhaustive set. From a methodological point of view, the FRBR model was developed on the basis of extant normative documents identifying the data elements that met each user task for each entity; would it not be better to take user tasks as a basis, and to list all data elements that would best suit those user tasks? That way, we could make sure that no data elements are missing in current normative documents. From the CRM perspective, the FRBR model has too many attributes for each entity and some attributes are not assigned to the appropriate entities: the original language of a work (only expression has language), medium of performance for a musical work (performance is only for expression), revisability of an expression (is of a work), and extensibility of an expression (is of a work).

Maja Žumer suggested therefore that we review all of the attributes defined in the FRBR Final Report, basing on user functions and specific needs within functions, and taking electronic resources more into account.

Discussion followed debating the differences in needs of a museum community versus a library community and the benefits of re-examining the library model to identify issues for re-examination. One of the benefits of a conceptual model is to be useful in different ways and the ability to apply it creatively. Identifying the source of the data was seen as very important. Serving the needs of the user was agreed as foremost objective. Having systems designed to take advantage of inherited attributes is a challenge as most systems now use flat files. There is great potential to share information across communities (get basic identifying information from publishers and augment that with information from libraries, etc.). The cataloger may not need to know if an element identifies a work or an expression, but just provide it.

At the end of Session B, there was a discussion about the notions of flexibility and ambiguity.

Ed O'Neill made the point that predictability is perhaps more important than consistency (some degree of inconsistency can be tolerable from the moment you can predict the kind of treatment you will find in a bibliographic record that was created by a given library).

Sally McCallum pointed out that aggregates are the biggest problem of inconsistency.

Ed O'Neill opined that there are many problems with omitted or missing information: "If you tell me that you omitted the illustrator and the foreword, then I can deal with that, but if you don't, I can't predict it." He admitted however that we have to worry about the costs of cataloguing.

Ketil Albertsen insisted that with millions of documents, as in the context of the Paradigma project, cataloguing has to be automated.

Godfrey Rust stressed that an important issue that should be addressed in the model is the question "Who says that?"

Someone said that increasingly libraries will be in a position of consuming data from various sources, and will have to evaluate the reliability of such data.

Jennifer Bowen opined that we should focus on the flexibility of FRBR; instead of keeping wondering: "Is this a Work or an Expression or a Manifestation?", we should focus on how to display it.

John Attig replied that some attributes are for the sake of collocation, and in that case flexibility works against us.

Session C: FRANAR/FRAR state-of-the-art and consequences for implementation and Subject access in FRBR

Glenn Patton: "FRAR: extending FRBR concepts to authority data"

The FRANAR Working Group is reviewing all entities from Group 2 and 3. The convergence of activities in 1998 with the publication of the MLAR document, the International Conference on National Bibliographic Services, and the FRBR call for extending the model to authority data. FRANAR has representatives from 7 countries and Tom Delsey was hired as a consultant. The terms of reference are to define the functional requirements of authority records, to explore the feasibility of an International Standard Authority Data Number (ISADN), and to liaison with other groups. As for the ISADN, it was proposed in the 1980s in GARE but declared unnecessary in MLAR in 1998. However, some people still feel the need for such a number and the FRANAR Working Group will address this issue yet again. The focus for authority information is a shift away from a single form of a heading for an entity used by everyone in the world to the use of multiple forms, depending on the needs of the user. The focus also is a shift away from sharing physical records to sharing the intellectual product that is in records. The Functional Requirements of Authority Records report is in draft and covers how authority files function, clarifies concepts to refine the process and practice of authority work for the future. The functions of an authority file are seen to 1) document decisions, 2) serve as a reference tool, 3) control forms of access points, 4) support access to bibliographic files, and 5) link bibliographic and authority files. The user tasks are: find, identify, contextualize, and justify. The entity "family" is reintroduced after discussions with the archival community, although one can declare family to be a type of corporate body. As a key concept it is noted that entities are bibliographic entities, that reflect the constructs of rules and what is perceived as a specific instance of an entity (which varies across rules). The authority files for archival communities do not directly parallel the library model and another difference appears in the rights management authority files where a work is defined differently. The FRAR report will suggest changes to FRBR definitions and changes to MLAR, GARR, and UNIMARC Authorities.

During discussion it was noted that the burden of having an ISADN is the very expensive process and the costs of a maintenance agency.

Marcia Lei Zeng and Athena Salaba: "Toward an international sharing and use of subject authority data"

Group 3 seems to be incomplete: Where is the Time notion? Where is the Process notion? The Event entity seems to be a combination of time and place. There should be a distinction between concrete concepts and abstract concepts. How could Ranganathan's facets (personality, matter, energy, space, and time) be introduced into Group 3?

Do we need further types of semantic relationships in thesauri? For instance: "overlap", "subfeature of", "parent", "sibling", "like"...

What authority data currently exist in an authority record? or What authority data should be included in an authority record?

The FRSAR Group (Functional Requirements for Subject Authority Records) will focus on Group 3. Its terms of reference read as follows:

  • "build a conceptual model of Group 3 entities within the FRBR framework (Entities in Group 1 and Group 2 can be used as the subjects of works; but further inclusion of them will depend on the outcomes of the work of the FRANAR Working Group);
  • provide a clearly defined, structured frame of reference for relating the data that are recorded in subject authority records to the needs of the users of those records; and
  • assist in an assessment of the potential for international sharing and use of subject authority data both within the library sector and beyond."

They will look into what should be in an authority record.

Diane Vizine-Goetz: "Subjects in fiction: the experience with WorldCat"

Tom Delsey (as quoted by John Attig) proposes that two entities be added: Form/genre and Time, and 2 further relationships: Work is an example of Form/Genre and Work covers/depicts Time and Place.

Diane Vizine-Goetz actually proposes that 3 further entities be added: Genre, Time, and Summary. She noted that current subject heading strings include not only subject information but also genre, form, time, and place. In her work for FictionFinder she led with summaries, then genre terms, settings (covers/depicts time/place), subjects, literary form. She suggested marketing studies would help re-order those elements. She found that FRBR mapped well for FictionFinder.

Maria Nasiłowska: "Precoordination in subject indexing systems and FRBR model"

The FRBR model links subjects to works. Nasiłowska reported on her research of LCSH, KABA (the Polish translation of RAMEAU) and reported that the precoordination of syntax was not useful. She stated the author/title naming of expressions or manifestations was impossible. She found problems with the Group 3 entities when they only have a subject function.The FRBR model helps us identify 3 problems with current subject headings:

  • Heterogeneous headings (they combine two notions of different nature, e.g. a person and a concept);
  • "Lying" headings (they convey a semantic ambiguity — e.g. in the subject heading "Dickinson, Emily (1830-1886) — Criticism and interpretation", "Dickinson, Emily (1830-1886)" actually refers to Emily Dickinson's works — or a syntactic ambiguity — e.g., the heading "United States" is required, whereas the actual topic is American citizens living in USA (as a consequence, when a geographic term is given it can sometimes reflect the place or the people of that place) — or a formal ambiguity — e.g., some form subheadings, such as "Women authors", do not refer to an actual form);
  • "Double" headings express only one concept in two subject headings.

She recommended using subject headings only for "content" (as in content vs. carrier) and to stop at the expression level, but noted that is a problem for subject relationships in the FRBR model, which are only at the work level. Further recommendations were to include as needed the place of a topic and the time of a topic as well as the form of a topic and to take the topic and form of a topic from authority files. Free-floating subdivisions should be replaced by topical headings. Form subdivisions that relate to carrier are actually already expressed somewhere else in the bibliographic record; form subdivisions that relate to the content should be studied into more detail. .She feels pre-coordination is dead.

During discussion it was pointed out that with post-coordination users will have to sort through many more irrelevant "hits." It was also pointed out that it is difficult to accommodate multiple controlled vocabularies. It was reported that studies in the 1920's in Germany revealed that both discipline and topics were needed and most useful in combination. Another speaker suggested also 'audience' and 'aspects' in which the categories are placed — that this also applies to summaries so they can be clearly labeled for users and machine-manipulation. It was suggested to exploit the distinctions but not expect a user to know those distinctions.

Session D: Implementation of FRBR

Trond Aalberg: "Formats and FRBR catalogues — where's our focus?"

Some attributes defined in FRBR are overspecialized, whereas they are not differentiated in MARC formats; conversely, some other attributes defined in FRBR are too generic, whereas they are expressed in MARC formats as a number of specialized fields.

Ketil Albertsen: "What do we want to identify? — FRBR and identifier semantics"

The Norwegian Paradigma project team needed to develop identifiers for harvested objects. But identifiers are not as simple as they seem. Ketil Albertsen listed more than 30 issues they considered while designing their identifiers, along with pros and cons, and an estimate of how sure they are of their decision: ID value carries no information about identified object; ID values should use a restricted character/symbol set; Check digits are included in ID display, etc. etc.

Patrick Le Bœuf: "Identifying textual 'works': ISTC: controversy and potential"

ISTC could be used in order to identify the Work notion in the sense defined by FRBR, provided the privileged relationship between a textual Work and the language in which it was created is recognized, either through the "Original language of the work" attribute as defined in FRAR, or through the "has representative expression" property as defined by the FRBR/CRM Harmonization Group, or by defining a new relationship, "was first expressed in" or "was conceived in", between the Work entity and some Language entity, as "language" is a mandatory element in ISTC metadata.

Thomas B. Hickey: "Exchanging FRBR information"

The FRBRization undertaken by OCLC was done primarily at the Work level. It is used in Open WorldCat and FictionFinder, and will be visible in FirstSearch displays this fall (2005). All the information used can be stored in MARC21 (although the output of the processing may not fit so well). MARC21 authority data served to map personal names using cross references and to map author-titles using cross references (which is especially useful for title variants, e.g. translations). "pKeys" are author-title keys for matching that were derived from MARC-like records and authority data. The xISBN Web service takes an ISBN as input and returns a list of ISBNs in the associated work.

Session E: Interaction with the library community and beyond

Barbara Tillett: "FRBR and Cataloguing Rules: Impact on IFLA's Statement of Principles and AACR/RDA"

Both the future cataloguing code Resource Description and Access and the IFLA New International Principles for Cataloguing are strongly influenced by FRBR concepts and the user tasks defined in FRBR.

Patrick Le Bœuf: "'Convergence is the Goal': Activity Report of the IFLA FRBR/CIDOC CRM Harmonization Group"

The FRBR/CRM Harmonization Group's activity will eventually result in an objet-oriented version of the FRBR model, which will account for the Event notion and will look more like an actual ontology.

Godfrey Rust: "Thoughts from a different planet (only slightly different)"

FRBR would benefit from an ontology, especially with the emphasis being placed on relationships.

The FRBR systems librarians are designing now will be mainstream in several years time. Is librarians' world going to get simpler or more complex? Can librarians predict the functional requirements they will have for 2010 or 2015? If not, then they should focus on the underlying flexibility of the model. Settle the model and put the change management into an ontology.

Librarians should use what is useful in other people's standards and metadata, and find ways of helping them make them better.

FRBR Review Group's Conclusions from the Workshop, as presented on May 4th

At the end of the Workshop, the FRBR Review Group presented the following conclusions:

Session A:

"Agglomerates": What is the Expression of an Agglomerate? The FRBR Revision Group will post on its Web site more examples for commentators to react, so that consensus can be reached about the best way to deal with such complex entities as anthologies, collected vs. selected works, serials, etc.

"Continuing Resources": The FRBR Review Group will contact the CONSER Task Force on FRBR & Continuing Resources in order to take benefit of their approach to continuing resources.

"Web Resources": That utmost complex topic is postponed for the time being.

Session B:

The FRBR Review Group will focus on the revision of attributes and relationships that are defined for Group 1 entities, as the FRANAR and FRSAR Study Groups will deal with Groups 2 and Group 3 of entities.

A "middle implementation model" will be developed by the FRBR Review Group; that "middle implementation model" will include a practical definition for the Expression entity.

Session C:

Is the Subject relationship a Work property only?

The FRBR Review Group thinks that "aboutness" is actually only at the Work level, but that subject headings such as we currently know them do not deal exclusively with "aboutness". The FRSAR Study Group will explore that.

Session D:

The FRBR Review Group will contact vendors* in order to ask them what they think the issues are for introducing the FRBR concepts into library catalogues. What do they expect from us in order to go ahead? What are their needs?

The FRBR Review Groups will besides strive to promote and stimulate research and study by doctoral students, professors, etc.

* Note: at the time this report is being written, this has actually already been done by Alan Danskin.

Session E:

Godfrey Rust's suggestion that the FRBR model would benefit from an ontology is accepted, and it is acknowledged by the FRBR Review Group that the FRBR/CRM Harmonization Group is going in that direction.