ANALYSIS OF KEY METADATA FOR IDENTIFYING DUPLICATES IN BIBLIOGRAPHIC RECORDS

Abstract

This study addresses the issue of duplicate bibliographic records in library information systems, a problem that is becoming increasingly relevant with the growth of digital catalogs. It specifically examines the key metadata fields used for comparing records and identifying duplicate entries. The analysis includes critical metadata fields such as title, ISBN, publisher, place of publication, publication date, pagination, series, and additional attributes used for identifying editions. Special attention is given to the variability of data within these fields, including issues arising from misplaced subfields (e.g., place of publication instead of year, or vice versa) and the use of various date formats, such as copyright dates, ranges, or approximate dates. The study explores the specifics of multi-level records, particularly for journals and multi-volume publications, as well as errors caused by data migration between different automated library information systems (ALIS). The research demonstrates that, despite the existence of ISBD, UNIMARC, and other standards, a significant proportion of inconsistencies persist in bibliographic records, complicating automated processing. Fields containing publishers and places of publication exhibit a high degree of variability, with unique values accounting for only 9% to 38% of total records. During data clustering using the nearest neighbor method with the Prediction by Partial Matching (PPM) algorithm, the number of unique values was reduced by 6–24%, highlighting the potential of automation to improve the efficiency of manual record editing. The findings make a substantial contribution to the development of effective approaches for enhancing automated bibliographic data management systems, optimizing duplicate detection, and improving the overall quality of library databases.