The SAC Subcommittee on Subject Relationships/Reference Structures was authorized at the 1995 Midwinter Meeting and appointed shortly before Annual Conference. Its creation was one result of a discussion of how (and why) to promote the display and use of broader-term subject heading references, and its charge reads as follows:
- the kinds of relationships that exist between subjects, the display of which are likely to be useful to catalog users;
- how these relationships are or could be recorded in authorities and classification formats;
- options for how these relationships should be presented to users of online and print catalogs, indexes, lists, etc.
By the summer 1996 Annual Conference, make some recommendations to SAC about how to disseminate the information and/or implement changes.
At that time assess the need for additional time to investigate these issues.
The Subcommittee's work on each of the imperatives in the charge was summarized in a report issued at the 1996 Annual Conference (Appendix A). Highlights of this work included the development of a taxonomy of 165 subject relationships; a demonstration that, using existing MARC coding, catalog systems could be programmed to generate references they do not currently support; and an examination of reference displays in several CD-ROM database products. Since that time, work has continued on identifying term relationships and display options; on tracking research, discussion, and implementation of subject relationships in information systems; and on compiling a list of further research needs.
Subject term relationships
In defining and classifying the subject relationships that exist within the English language, the Subcommittee hoped not only to demonstrate the wide variety of these relationships, but also to begin development of a knowledge base for future improvements in subject retrieval. Such improvements range from a more logical syndetic structure for Library of Congress Subject Headings to applications in concept mapping, information visualization, data mining, and artificial intelligence.
Subcommittee members reformatted the taxonomy of 165 relationships which Dee Michel had compiled from various sources
Appendix B, part 1) into a hierarchical display (
Appendix B, part 2). Each member of the Subcommittee then tried to characterize the see-also references from the LCSH term "Family" using this taxonomy. The experience highlighted limitations of the taxonomy as a working tool (overlapping categories, not enough examples, overall length) and problems with the LCSH example (one term embodying two separate concepts, varying closeness of relationships). In a follow-up exercise, pairs of subcommittee members were assigned six LCSH terms apiece and asked to look for clusters of similar relationships among the see-also references. The Subcommittee used the results of this exercise along with the taxonomy to develop a checklist of subject relationships suitable for use in information retrieval (
Appendix C).
We found a great variety in semantic relationships. A lengthy but not exhaustive literature review identified over 100 associative relationships, over 50 equivalence relationships and over 20 hierarchical ones. This variety was expressed in depth as well as breadth. When organized in terms of increasing specificity, the tree was five levels deep, e. g.:
Equivalence/same lexical item variants/morphological variants/inflectional suffix variants/irregular plural-singular pairs
WOMEN/WOMAN
Hierarchical/inclusion relationships/physical inclusion/whole-part/concrete whole-part/anatomical whole-part/anatomical system whole-part
NERVOUS SYSTEM/CENTRAL NERVOUS SYSTEM
Associative/pairs from same hierarchy/causal relationships/dependency relationships/entity-precursor relationships
RAIN/CLOUD
One's first reaction is that it is not necessary to encode semantic relationships as specific as these. But on what basis can we make that judgement? One could imagine a system that would automatically recognize regular plural terms in English and singular terms as equivalent. Wouldn't it make sense to encode at least the irregular ones? Anatomical whole-part is considered significant enough to be mentioned by two or three writers. And wouldn't thinking about entity and precursor be a good way for catalogers establishing a new index term or subject heading to generate useful RT's?
Display options
Appendix D provides an evaluative report on the treatment of subject references in the CD-ROM indexes (Reader's Guide, Silver Platter's MLA Bibliography and Medline, and RILM) examined earlier by the Subcommittee. It demonstrates how an emphasis on simplicity in programming over clarity of presentation can render references almost as unusable as if they were not displayed at all. In the process it highlights the value of such features as clear explanation and labeling, sophisticated list ordering to bring out the meaning in subdivided headings, and the display of thesaurus relationships by category.
One pair of Subcommittee members used the clustering exercise mentioned above to develop a conceptual (i.e. faceted or categorized) display of the LCSH terms and references they analyzed (Appendix E). A preliminary survey of subject referencing in OPACs (described in Appendix F) found that most libraries display cross references above the bibliographic records in a line-by-line list, but that some catalogs present as a single choice the option of viewing references. In Web-based catalogs, each reference is a clickable link to the bibliographic records using the heading, as well as to its set of related terms.
Status of research and implementation
The Subcommittee noted that subject relationships and referencing play a role in a variety of prototype information retrieval systems. Some, such as those developed by Karen Drabenstott and her colleagues, use reference displays as part of a suite of tools for managing large search results. Others employ concept mapping and information visualization techniques to present subject relationships in treelike and maplike displays; examples of these may be found at the Project Aristotle(sm) and The Big Picture(sm) clearinghouses maintained by Gerry McKiernan at www.public.iastate.edu/~CYBERSTACKS/Projects.htm.
Just within the past twelve months many libraries and vendors have begun to add a subject referencing feature to OPACs. A preliminary survey of Web-based and conventional online catalogs at the end of May turned up 86 of 120 conventional, and 35 of 50 Web-based OPACs offering some sort of subject referencing (around 70% of each group). A separate survey by vendor found 14 of 19 conventional catalog systems, but only 6 of 17 Web-based systems, offering references to narrower, broader, or related terms. The survey showed widespread referencing to narrower and "related" terms, as well as a small but significant number of libraries offering broader-term references too. It also suggested, however, that larger catalogs may be much less likely to offer subject referencing than medium-sized and smaller ones, and that systems considered "experimental" or "cutting-edge" are especially unlikely to offer subject referencing based on a controlled vocabulary or a syndetic structure. (The report of this study comprises Appendix F.)
Unfortunately, many vendors and librarians alike remain unconvinced of the usefulness of subject referencing in OPACs. A position paper addressing this issue is included in the Subcommittee's report as Appendix G.
Scope of the charge—and of the subcommittee's work; research needs
The topic of subject relationships and reference structures in information systems is broad in scope, with a number of aspects the Subcommittee did not have time to explore. These include:
- How subject references are coded in non-MARC databases
- Thesaurus-based searching and other "hidden" applications of subject relationships
- The Art and Architecture Thesaurus (though we did work with AAT staff in developing our checklist)
- The larger search-assistance environment under development, and its use of subject referencing (but with due regard to Karen Drabenstott's research)
- The use of subject relationships in classification schemes
Other issues, including equivalence relationships, GUI displays and the determination of cost effectiveness, were considered but not explored in depth. The twin emphases of our work were spotlighting the range and variety of existing subject relationships, and building a case for making explicit subject referencing available in catalogs and other information systems.
The Subcommittee's investigations did yield a number of topics for additional research. These issues relate primarily to the implementation and use of subject relationships in library catalogs and cover the following areas: subject authority control tools and standards, interface/display issues, encoding issues, cognitive research, feasibility issues, maintenance, and current systems. The list of research topics comprises Appendix H. Suggestions for further reading, arranged both alphabetically and by subject categories, can be found in Appendix I..
Findings and recommendations
Based on its investigations of the issues set forth in its charge, the Subcommittee offers the following findings and recommendations:
- Identification of subject relationships
Finding: A great variety in terms of depth and breadth was identified (taxonomy). A subset was identified as being of potential use and is presented as a checklist.
Recommendation: Libraries should use relationships in the checklist when creating references for new headings.
- Coding of subject relationships
Finding: Coding compatible with MARC formats can be developed to display relationships based on the checklist.
Recommendation: The library community, working through MARBI, should develop coding to support displays of all relationships in the checklist.
- Display of subject relationships
Finding: Many systems do not display currently encoded relationships (BT, NT, RT).
Recommendations: All systems should display currently encoded relationships. All systems should expand these relationships to include the more specific ones proposed in the checklist.
Finding: Alphabetical displays under highly referenced terms result in long lists of relationships, which seems ineffective.
Recommendation: References should be clustered based on relationships in the checklist.
- Research
Finding: Not much is known about the usefulness of subject relationships in information retrieval.
Recommendation: Research should be done on such issues as user behavior, cognitive aspects of database searching and navigation, comparative performance of semantic and other types of data relationships, and the advantages and disadvantages of different displays.
- Further issues
Finding: Other issues were raised that this subcommittee did not have time to investigate in detail.
Recommendation: SAC should establish another subcommittee (or two) to study these issues, i.e.:
- the use of relationships by users vs. catalogers
- cost/benefit analysis of this Subcommittee's recommendations
- other aspects of semantic relationships, e.g. in classification, subdivisions, scope notes
- expansion of definition of RT relationship by the Library of Congress
- equivalence relationships
- an ALA program on subject relationships and reference structures
SAC Subcommittee on Subject Relationships/Reference Structures
Jane Greenberg, University of Pittsburgh
Harriette Hemmasi, Rutgers University
Patricia Kuhr, H. W. Wilson Company
Dee Michel, University of Wisconsin
Steven Riel, Harvard University
Gary Strawn, Northwestern University
Gregory Wool, Iowa State University, Chair
Lynn El-Hoshy, Library of Congress, Consultant
APPENDIXES
- SSRRS Report to SAC, July 1996
- Taxonomy of Subject Relationships:
(part 1) Alphabetical Display
(part 2) Hierarchical Display
- Checklist of Candidate Subject Relationships for Information Retrieval
- Review of Reference Displays in Selected CD-ROM Abstracts and Indexes
- Analysis of Relationships in Six LC Subject Authority Records
- Report of a Preliminary Survey of Subject Referencing in OPACs
- LC Subject Referencing in OPACs--Why Bother? (position paper)
- Research Needs on Subject Relationships and Reference Structures in Information Access
- Bibliography on Subject Relationships
REFERENCE
Bates, Marcia J. "Subject access in online catalogs: A design model." Journal of the American Society for Information Science 37(6):357-376; Nov. 1986.
