return to curriculum vitae

Dagobert Soergel
College of Information Studies, University of Maryland

Improved access to the LC Classification and the LC Subject Headings for online searchers and catalogers through conceptual analysis
Presentation at OCLC, March 14, 1989

Classifications, such as LCC and DDC, and subject heading lists, such as LCSH are not designed for online searching and are very difficult to use, but they are here to stay. They provide a major avenue of subject access to huge collections searchable through online public access catalogs.

This paper proposes a system which would enable a user to formulate a query in elemental concepts chosen from a faceted classification and find documents likely to be relevant to her query based on the LC class number and subject headings assigned; the system would allow postcombination searching, which is best suited for online systems, on the basis of precombined descriptors, which are meant for manual systems. The system would provide an assist mode, in which it would guide the searcher through a facet analysis of the search topic. It would let the searcher display the class numbers and subject headings found before searching for documents. With a few additions the system would be useful for catalogers as well. The cataloger would specify the document topic as a combination of elemental concepts, and the system would suggest LC class numbers and subject headings for consideration. The system would also guide the cataloger through LCC tables using a menu format and compute the final class number from the menu choices.

The knowledge base for such a system includes the following:
  1. A universal faceted classification of elemental concepts.
  2. For each class and each subject heading a combination of elemental concepts that expresses it.
  3. The rules for using tables in LCC and for combining main headings and subheadings in LCSH.

The paper describes briefly how the system would use this knowledge.

The major task in developing such a system is constructing these components of the knowledge base. The paper details a procedure for this huge task. This procedure relies as much as possible on computer assistance, using existing faceted classifications as a starting point and exploiting two facts:

  1. The conceptual deep structure of a class or subject heading is often reflected in the linguistic surface structure of the class name or title or the text of the heading.
  2. A subordinate class inherits the elemental concepts that express its superordinate class.

The procedure also employs the rules for using tables in LCC to "unpack" class numbers that result from using tables. The task also requires a substantial amount of intellectual work; the procedure provides for optimal division of labor between the work of a computer program and the work of the human editors.

top of page