DORS: DDC Online Retrieval System
The use of traditional classification as interfaces to online catalogs is still a matter for discussion and experimentation. While interest in the matter dates from the 1960s, with the ground-breaking work of Freeman and Atherton, and has continued sporadically, most of the work in this area has been of a speculative and illustrative nature. Not until recently has it been possible to experiment with large prototype systems. What has made this possible has been the rendering into machine-readable form in 1984 of the Dewey Decimal Classification (DDC). The most well-known of the DDC prototype systems was that developed by Markey et al. at the OCLC Online Computer Library Center in 1986. Another system based on the machine-readable DDC is currently being developed at OCLC under the direction of Diane Vizine-Goetz. This system, whose purpose is to support online classifying, offers a variety of functions, including keyword searching, hierarchical browsing, and multiple display options. A third prototype system using the machine-readable DDC has been developed at the University of California, Los Angeles (UCLA), and is reported upon in this paper. It differs from the OCLC system in that it is designed primarily for catalog users other than classifiers. The distinguishing feature of the system is an automatically generated chain index. Following a discussion of the specifications deemed desirable in a DDC catalog interface, the system that was developed at UCLA to conform to them is described.
A classification system like the DDC can be regarded as though it were a language, consisting of a vocabulary and a semantic structure for organizing and displaying vocabulary elements in juxtaposition to one another. The fact that such systems have vocabularies recommends them for the purpose of enhancing the entry vocabulary of an online catalog. The Market et al. project has demonstrated how the DDC "Relative Index" and schedule captions can be used to augment searching vocabularies.
The first requirement, then, for a classification interface is that it enhance searching vocabulary. However, were such an interface to be developed solely for this purpose, questions of effectiveness and cost could be raised. Terms from the index of a classification might be good search terms; however, terms from schedule captions are often too general or ambiguous to be useful for this purpose. More to the point, however, is that introducing the complicated mechanism of a classification simply for vocabulary enhancement would not be cost-effective. Other means to this end could well be cheaper and more effective, for instance, making terms from a book's table of contents searchable.
The attribute that most distinguishes a classified or systematic subject approach to information from an alphabetic subject approach is the particular kind of structure it imposes upon vocabulary terms. Classifications cluster the vocabulary associated with a concept at different levels of specificity. Thesauri exert a similar type of vocabulary control, and some thesauri, like classifications, even display terms in hierarchical displays. Classifications, however, go beyond thesauri by semantically structuring not only the vocabulary associated with concepts but also the concepts themselves. Classifications have sometimes been likened to semantic nets, in which concepts are linked by meaning relationships. They have also been likened to knowledge trees, in which each concept is comparable to every other concept, in the sense that its position in the scheme is defined with respect to every other position. While thesauri consist of hundreds, perhaps thousands, of term clusters, classifications attempt to integrate these clusters into meaningful, monolithic wholes. For this reason, classifications can be said to partake more of the nature of a knowledge base than do thesauri. …