What Does Transaction Log Data Tell about Collection-Level Subject Access? Oksana L. Zavalina, Assistant Professor University of North Texas College of Information, Department of Library and Information Sciences [email protected] Background Research Questions Studies of subject access to date have focused on itemlevel information discovery. They have shown that users experience problems due to quality and application of metadata. Method: Transaction Log Analysis A. To what extent do the users explore aggregations by subject? Aggregations of digital collections provide a significant source of data for analyzing collection-level subject access and the role of collection-level metadata. Opening History http://imlsdcc.grainger.uiuc.edu/history IMLS-funded national aggregation of almost 900 cultural heritage digital collections B. How does collection-level information seeking behavior differ from item-level information seeking behavior in aggregations? • Systematic sample (12 weeks throughout 1 year) C. How does collection-level metadata help the users of aggregations find digital objects and collections of objects through searching? • Retrieval sets for 501 unique collection search queries • Collection-level and item-level user interactions (query-level analysis) • FRBR set of entities adapted as analytical framework. Findings Findings 1. Almost even % of collection and item interactions 2. Prevalence of subject browse 6. Role of free-text and subject-specific collection metadata fields in matching user search queries Percentage of collection searches that retrieved results with: 93% 74% match in a field 50% Description field (match in 93% of searches) Subjects field (50%) Other subject-specific fields (3-12%) Objects/genre Geographic Coverage Temporal Coverage. 3. Collection metadata records viewed 4 times as often as item metadata records 4. Collection-level searching is mostly subject searching FRBR Group 3 or subject entities: object (e.g., “cookbooks”) place (e.g., “Central Florida”) concept (e.g., “transportation”). 5. Similarities and differences in collection and item searching • Prevalence in subject searching (FRBR Group 3) • More person, event, ethnic group, and class of persons searching at item level. 36% Distribution of search categories in: collection-level search 28% Free-text Description fields 21% of collection records would not be retrieved without them 27% 26% 24% item-level search 7. Role of free-text and subject-specific collection metadata fields in discovery of digital collections 22% 19% 17% 27% 12% match in a field 25% 3%4% 3% 7% 3% 2% match only in this field 24% 21% 15% 11% 7% 12% 8% Percentage of collection records retrieved with: 36% 13% 10% 8% 8% 9% 7% match only in this field Subject-specific controlled-vocabulary metadata fields overall, 41% of collection records would not be retrieved without them. 12% 13% 9% 5% 93% of user searches are matched by Description field in at least 1 collection record; search term(s) can be matched only through Description in 74% of searches. Conclusions and implications Future research 1. 2. 3. 4. • Longitudinal comparative analysis of transaction log data from various aggregations of different size, scope, and focus. • Session-level analysis of log data. Users of aggregations often search and browse at collection level; view collection records. Subject searching and browsing are prevalent among collection-level user interactions. Provision of collection-level metadata in aggregations of digital collections is important for subject access. Free-text collection metadata in aggregations should be complemented with subject-specific controlled-vocabulary collection metadata fields.
© Copyright 2026 Paperzz