Dissertation Information for Judith Wood NAME: - Judith Wood
- (Alias) Judith Barlow Wood - (Alias) Judith B. Wood
DEGREE:
- Ph.D.
DISCIPLINE:
- Library and Information Science
SCHOOL:
- Case Western Reserve University (USA) (1983)
ADVISORS: - William Goffman
COMMITTEE MEMBERS: - Thomas Gleason Morris - William M. Shaw Jr. - Adel Mahmoud
MPACT Status: Fully Complete
Title: A method for comparing the structure of a discipline to the structure of a bibliographic file for that discipline
Abstract: The less than optimum satisfaction with present information and bibliographic files and their associated retrieval systems suggests that the structure of these files may not closely reflect the structure of the disciplines those files represent. This study presents a method for objectively and quantitatively examining and comparing structures of disciplines with structures of the bibliographic or information files of those disciplines. It tests the hypothesis that present file structures are not highly correlated with the structures of the day-to-day operational world dealt with by the end users of those files.
A structure is derived for the discipline of tropical medicine based on the signs and symptoms of tropical diseases. The National Library of Medicine's MEDLINE file is examined for a corresponding structure, based on the use of the same signs and symptoms as index terms for articles on the tropical diseases. For the discipline, the structure consists of intercommunication classes resulting from the relationships between the signs and symptoms in the diseases. For the file, the structure consists of the intercommunication classes resulting from the interrelationships of these signs and symptoms as index terms. The classes are formed by setting up matrices showing the probabilities of relatedness of the signs and symptoms in each set. Class membership is determined at various thresholds by applying the single-link clustering technique.
Overall centrality numbers are calculated for each sign and symptom in both sets using Goffman's Digraph Measure. The signs and symptoms of each set are ranked by their centrality numbers, and the two sets compared. Both the Spearman Rank Order Correlation and the Pearson Product Moment Correlation procedures reveal a positive but low correlation coefficient (0.315 and 0.316, respectively), as hypothesized. The actual correspondence is probably even lower since appropriate MEDLINE subject headings could be found for only 58% of the signs and symptoms. Implications of this low correlation are discussed, and suggestions are made for further studies. Addition of sign/symptom check tags to the MEDLINE system is recommended.
|