Supplementary Data for \"Full-text-validated Paired Audit of Bibliographic Metadata in Scopus and Web of Science Using a Matched Corpus of 2,141 North Korea-affiliated Publications, 1976–2024\"

This dataset accompanies the article \"Full-text-validated Paired Audit of Bibliographic Metadata in Scopus and Web of Science Using a Matched Corpus of 2,141 North Korea-affiliated Publications, 1976–2024\" (Hong, Park, Kim, Kim, Yun, & Cho, 2026). It contains the validated coding framework, paired comparison tables, and adjudication rules used to audit four core metadata fields—digital object identifier (DOI), author affiliations, corresponding authorship, and reference counts—in Scopus and Web of Science across a matched corpus of 2,141 North Korea-affiliated science and technology papers indexed in both databases from 1976 to 2024. The dataset includes: (1) paired error-coding tables for each of the four metadata fields; (2) the manually verified reference-count set (v-set) derived from source-document inspection of 2,127 papers; (3) 2 × 2 contingency tables underlying the McNemar tests reported in the manuscript; (4) aggregated match-rate and error-rate data underlying Figures 1–6; (5) coding rules and adjudication criteria applied during error classification; and (6) inter-coder reliability assessment results based on 100 stratified sample papers (Cohen's kappa = 0.95; 99.3% agreement across 800 coding decisions). Source PDFs of the audited publications are not redistributed because they remain subject to publisher copyright and licensing restrictions. Individual author names are aggregated or coded where appropriate to avoid providing consolidated personal-name lists.