The library community has made various efforts to address the problem of using MARC and FRBR together and adopted several approaches to interoperability between these different structures. Delsey mapped MARC21 elements to FRBR attributes to achieve interoperability.³ Aalberg refined the FRBR attributes and created mapping tables to match FRBR attributes to MARC elements.⁴ Both focused on similarities between the two sets of descriptive elements. However, such efforts have indicated that many FRBR attributes cannot be mapped directly to MARC and vice versa. More importantly, both Delsey and Aalberg attempted to map between MARC elements and FRBR attributes without considering either the structural differences in their architectures or the procedural differences that exist between the two systems. MARC's single-layered structure and FRBR's multilayered hierarchy affect the semantics of their components differently: a component in either scheme might have a specific meaning that influences or is influenced by the relational structure of the standard. The weaknesses of approaches such as those of Delsey and Aalberg are rooted in the mapping method they have used. According to Kurth, Ruddy, and Rupp, mapping is the process of establishing relationships between semantically equivalent elements in different structures.⁵ Thus mapping refers to the process of associating elements from one set with elements from another set, thereby providing conceptual connections between data elements in two or more bibliographic structures. However, establishment of relationships at the structural level is necessary to achieve full interoperability between two or more schemes. Simple mapping is not sufficient because it identifies relationships only at the level of the data element.

Riva adopted a detailed bidirectional mapping between FRBR relationships and MARC linking entry fields.⁶ In this mapping, Riva used Tillett's taxonomy of bibliographic relationships, which has seven major classes: equivalence, derivative, descriptive, whole-part, accompanying, sequential, and shared characteristic. Through these divisions, Riva mapped each MARC linking entry field to one or more entries in the FRBR relationship. The mapping focused on differences in scope and level of detail represented in the categorizations of bibliographic relationships, and it provided clear relationships between MARC21 fields and FRBR relationships. Although Riva's mapping allowed direct links between closely related works, the scope was limited to MARC 76X-78X fields and did not provide comprehensive mapping because of differences in granularity between MARC21 and FRBR.

Monch and Aalberg proposed a prototype system for the automatic extraction of FRBR model entities from MARC records.⁷ This extraction required mapping between MARC fields and FRBR entities. For the mapping, they used the framework BIBSYS, which is based on the MARC format, to convert MARC fields and subfields to FRBR entities given the information stored in a set of existing MARC records. In mapping MARC fields to FRBR entities, Monch and Aalberg introduced what they called an attribute layer to extract consistent information from the records. Using the attribute layer, information from MARC records could be mapped to a set of selected FRBR entities. Although this approach integrated MARC fields and FRBR entities based on their generic meanings, it required a new framework to merge one scheme with another. Even though this approach eliminated heterogeneity between MARC and FRBR through construction of a new framework, it could not retain the entire structure or characteristics of either MARC or FRBR.

Each of these approaches sought to eliminate heterogeneity between MARC and FRBR. However, the weaknesses of these approaches are obstacles to achieving full interoperability. One main reason for their failures is that these approaches focus on the similarity of names to establish relationships between elements in the two structures, frequently resulting in the identification of ambiguous, misleading, or inaccurate relationships that present significant obstacles to achieving interoperability between MARC and FRBR.

Because MARC and FRBR are bibliographic systems with very different structures, the authors sought to achieve interoperability between them and to fully capitalize on the advantages of each by constructing a conceptual framework that would be flexible enough to cope with the complexity of digital resources while functioning as a mediator between the two systems. To construct this conceptual framework, the authors analyzed MARC elements and FRBR entities and attributes in light of their respective structures. They then categorized elements and attributes according to their intended application or use. Only those elements that described resources (MARC) or indicated bibliographic relationships (FRBR) were analyzed, and elements that had the same or similar meanings in each structure were categorized. This provided a basis for mapping elements from each structure; MARC elements in one category were mapped only with FRBR attributes and entities in the equivalent category.

While categorizing elements, the authors identified four types of matching and used them to group elements: exact matching, analogous matching, partial matching, and nonmatching. On the basis of analysis of semantic relationships between categories, the authors identified four levels of nesting for the proposed structure: main class, class, subclass, and instance. Finally, the authors merged the mapped elements to form a new structure that was based on the semantic relationships between elements. This approach capitalizes on the strengths of mapping and merging to achieve interoperability between two heterogeneous structures.

Traditionally, cataloging consists of three core elements: author, title, and subject. Although evolution of the cataloging environment has added additional elements (e.g., physical description), the core elements are critical for describing resources. Theoretically, every resource has an author who is responsible for the intellectual content of the resource, a title that represents the resource, and a subject or topic that the resource addresses. In addition, the content is carried by a specifiable medium, be it the printed page or a digital website. A bibliographic record is not “complete” without the author, title, subject, and description elements. Accordingly, the authors began construction of the conceptual framework by extracting core bibliographic elements from both MARC and FRBR. After extraction, they categorized these elements according to their representational use to provide a basis for linking between MARC elements and FRBR entities and attributes.

Following categorization and the mapping of MARC elements with FRBR attributes, the authors constructed a conceptual structure that would function as a mediator between MARC and FRBR. This structure, which consists of main classes, classes, subclasses, and instances, is made up of abstract elements and is not intended to address actual descriptive elements in MARC, FRBR, or any other bibliographic standard.

Main classes are groups of identical or similar elements and attributes from both MARC and FRBR and correspond to the seven core categories derived during the categorization process. In both the MARC and FRBR systems, the elements and attributes in a main class are frequently distributed across the respective system. For example, both the 245 and 505 fields in MARC represent the title of a resource, but these fields are separated in the MARC system. FRBR separates attributes related to the title of a resource (title of work, title of expression, and title of the manifestation). In the conceptual structure, each main class provides a space in which these “distributed relatives” are brought together. In addition, any duplication of elements is eliminated by the grouping of identical or similar elements and attributes.

Among the types of match used in the mapping process, exact matches indicate that MARC and FRBR use the same elements to describe a particular aspect of a bibliographic entity. The element(s) and attribute(s) whose relationship is based on an exact match are used to form a class of data elements that connects MARC elements directly to FRBR attributes. Only those elements and attributes that share a comprehensive or identical referent were selected to form a class.

Each class is nested under its relevant main class (table 7). Analogous matches (those matches indicating that a similar but not identical meaning or referent is shared between a MARC element and a FRBR attribute) form subclasses. Although an analogous match does not indicate a comprehensive meaning encompassing the meaning of each subclass, it can demonstrate conceptual associations across a range of elements. Partial matches are identified as instances of classes and subclasses, with each instance representing the specific details of a bibliographic record. The full conceptual structure is provided in appendix B.

Main classes, classes, subclasses, and instances are nested hierarchically in the proposed conceptual structure, with the main class defining the semantic range of its subordinate classes, subclasses, and instances. A class brings together related elements under each main class and subclasses. Instances represent the more detailed aspects of a particular class and serve as the actual medium for connecting MARC elements and FRBR attributes within the conceptual structure.

The proposed conceptual structure is not intended for the description of specific resources but as a set of bibliographic data elements that can be linked directly both to MARC elements and to FRBR attributes. If an element in the proposed conceptual structure connects MARC elements with FRBR entities and attributes, MARC can be used for detailed descriptive elements and FRBR for the representation of bibliographic relationships. In this way, an element in the conceptual structure can be used to indicate both detailed descriptive elements and bibliographic relationships.

This structure also addresses some of the weaknesses in both MARC and FRBR. To overcome the lack of bibliographic relationships resulting from MARC's flat structure, MARC elements can be supplemented with FRBR's bibliographic relationships through connections established in the conceptual structure. In similar fashion, the lack of sufficient descriptive attributes in FRBR can be addressed by incorporation of related MARC elements.

The authors constructed a conceptual structure that can function as a mediator between the heterogeneous bibliographic systems MARC and FRBR. MARC is limited in describing information resources because of its rigid and single-layered linear structure. Although the MARC format is suitable for representing more traditional resources, such as books and print materials, weaknesses inherent in its structure prevent full representation of the complex nature and semantics of digital resources. In contrast, the FRBR model focuses on the organization of entities and attributes supporting multiple relationships between bibliographic entities. While FRBR can support representation of the multilayered characteristics of resources, it does not provide sufficient descriptive elements to fully represent bibliographic entities. Additionally, because the relationships between attributes in FRBR are predetermined, they might restrict the flexibility necessary to describe the dynamic nature of digital resources.

The conceptual structure proposed here serves as a mediator between MARC elements and FRBR attributes. In this conceptual structure, data elements are able to connect MARC elements and FRBR entities and attributes because the components of the conceptual structure were extracted directly from the core elements of both systems. Although the conceptual structure is not intended for the actual description of resources, it does provide a set of core bibliographic elements that, in association with an explicit structure of relationships, can make up for the weaknesses of both MARC and FRBR. More importantly, the elements of the MARC format can be used in association with bibliographic relationships in the FRBR model through the connections provided in the conceptual structure. In addition, systems that use MARC bibliographic records can be expected to demonstrate enhanced capability for information retrieval when combining, through the conceptual structure, the detailed description of MARC with the explicit relationships of FRBR.