TOWARD A COMPUTER-GENERATED SUBJECT VALIDATION FILE - FEASIBILITY ANDUSEFULNESS

Responding to the fact that the library community has long recognized the need for improved efficiency and reliability in subject authority control, we explored the feasibility of automatically creating a subje ct heading validation file by scanning the OCLC Online Computer Librar y Center, Inc. Online Union Catalog (OLUC). The premises were, first, that although the file would not he exhaustive, it would contain the m ajority of frequently used headings, and second the predicted level of accuracy in the file would be high. In approach, Ice focused on findi ng the density and distribution of assigned headings and the relations hip, if any, between density and error rate. We analyzed a sample file of Library of Congress-assigned headings from the OCLC Subject Headin gs Corrections database. The results of the study showed that (I)frequ ency of use and number of headings at a given rate of use are in inver se relationship; (2) a small number of headings with high frequencies of use accounts for a majority of total use, while a large proportion shows very low frequency of use; (3) topical headings account for abou t two-thirds of assigned headings; and (4) error and obsolescence rate s are both low, and both are in inverse relationship to the frequency of heading use. We concluded that an automatically generated subject h eading file is indeed feasible. Such a file would be useful for variou s purposes: to verify subject heading strings constructed by cataloger s, to update subject headings in catalog maintenance, and to validate subject headings during retrospective conversion.