Dissertation Information for Rao AluriNAME:
DEGREE:
DISCIPLINE:
SCHOOL: ADVISORS: COMMITTEE MEMBERS: MPACT Status: Fully Complete Title: SUBJECT ACCESS TO CATALOG RECORDS IN LARGE BIBLIOGRAPHIC DATA BASES Abstract: The Library of Congress subject headings provide the primary means of subject access to catalog records in large data bases such as OCLC. The retrieval performance of the LC subject headings is dependent on the size of the data base. Heretofore, no attempt was made to examine this relationship between the size of a bibliographic data base such as OCLC and the retrieval performance of LC subject headings. The present investigation, consequently, set out to examine the distribution characteristics of LC subject headings within the catalog records of the OCLC data base; and, based on this distribution pattern, to predict their retrieval performance when the data base consists of 10 million records. The retrieval performance of subject headings is measured in terms of (a) the number of postings per distinct subject heading, (b) the probability of retrieving up to an arbitrary number of bibliographic records; (c) the number of hits per query; and (d) the entropy of subject headings. Data for this project was collected from a sample of 33,455 catalog records from the OCLC data base. Retrieval performance measures of LC subject headings were computed by means of a negative binomial distribution model. This model assumes that the distribution of bibliographic records over subject headings follows the negative binomial probabilistic model. The analysis of the data indicates that there are 47,036 LC subject headings in the sample of 33,455 catalog records making the average number of subject headings per record as 1.4. Further, nearly 60% of the records contain, at most, only one subject heading. The retrieval performance measures indicate that, when the data base contains 10 million records, there will be 45,000 unique topical main headings and nearly 14,000 unique subdivisions. However, these main headings and subdivisions are combined to form 322,000 unique full topical subject headings. The number of bibliographic records associated with each unique complete topical heading would be 31 while the projected number of hits per subject heading, (assuming that the greater the number of postings associated with the subject heading, the greater would be the probability that the user would choose the subject heading to retrieve bibliographic records) would be 641. In contrast to full topical headings, the retrieval performance of main topical headings would be worse as the number of postings per heading would increase to 213 and the number of hits per subject heading to 11,516. Based on the retrieval performance of full topical subject headings, it can be concluded that full LC subject headings continue to perform well even when the data base reaches the size of 10 million bibliographic records. However, the users need to be able to identify full subject headings which satisfy their information requirements before retrieving bibliographic records. A new subject access technique, called Indirect Search method, is proposed which would assist the user of the subject access system in this regard. |
MPACT Scores for Rao AluriA = 0 Advisors and Advisees Graphgenerating graph, please reload |
Students under Rao Aluri
ADVISEES:
- None
COMMITTEESHIPS:
- None