Investigating the use of metadata record graphs to analyze subject headings in the digital public library of America

Purpose - This study furthers metadata quality research by providing complementary network-based metrics and insights to analyze metadata records and identify areas for improvement. Design/methodology/approach - Metadata record graphs apply network analysis to metadata field values; this study evaluates the interconnectedness of subjects within each I nub aggregated into the Digital Public Library of America. It also reviews the effects of NACO normalization simulating revision of values for consistency and breaking up pre-coordinated subject headings to simulate applying the Faceted Application of Subject Terminology to Library of Congress Subject Headings. Findings - Network statistics complement count- or value-based metrics by providing context related to the number of records a user might actually find starting from one item and moving to others via shared subject values. Additionally, connectivity increases through the normalization of values to correct or adjust for formatting differences or by breaking pre-coordinated subject strings into separate topics. Research limitations/implications - This analysis focuses on exact-string matches, which is the lowest-common denominator for searching, although many search engines and digital library indexes may use less stringent matching methods. In terms of practical implications for evaluating or improving subjects in metadata, the normalization components demonstrate where resources may be most effectively allocated for these activities (depending on a collection). Originality/value - Although the individual components of this research are not particularly novel, network analysis has not generally been applied to metadata analysis. This research furthers previous studies related to metadata quality analysis of aggregations and digital collections in general.