Coincidence of user vocabulary and Library of Congress subject headings: Experiments to improve subject access in academic library online catalogs.

摘要

Experimental research was needed to determine which retrieval processes in library online catalogs will increase the match success rate between user subject vocabulary and Library of Congress subject headings. Random samples of user subject terms from Northwestern University Library's NOTIS/LUIS online catalog were collected from transaction logs during the 1986/87 academic year. The user terms were described numerically, then compared to Library of Congress subject headings to determine the degree of match success.;Twenty-two experiments were conducted on match failures to determine how well modifications in search retrieval processes or file augmentation improved the match success rate in a simulated large academic library online catalog. Four categories of retrieval procedures were tested: (1) processing and correcting user input (i.e., spelling checker/dictionary, simultaneous singular and plural word form retrieval, and phonetic spelling); (2) searching the database (i.e., right truncation, string, and keyword searching operations in both the descriptor field only and full bibliographic records); (3) augmenting the database with authority files (i.e., subject, personal, corporate, and geographic name authorities); and (4) searching the augmented database (i.e., right truncation, string, and keyword searching applied to the augmented database).;The experimental results showed that the retrieval processes of right truncation, string searching with adjacency, and keyword searching with the Boolean operator significantly improved match success both with and without augmentation of the database with subject and name authorities and with the full text of bibliographic records. Neither subject nor name authority files improved match success if the online catalog required users to exactly match the Library of Congress form of entry.;The descriptive results showed that forty percent of all user terms exactly matched Library of Congress subject headings; user terms were short, with means of fewer than two words and thirteen characters per user term; users entered singular words eighty percent of the time; more than ninety percent of user terms contained no capitalization or punctuation of any kind; and more than one-fourth of all user terms contained a personal, corporate, or geographic name, but only one-third of these searches exactly matched Library of Congress subject headings.