East Asian UNESCO Intangible Cultural Heritage Metadata Audit Dataset and Reproducibility Code

Overview

This record provides an author-generated dataset and reproducibility code for auditing UNESCO Intangible Cultural Heritage (ICH) metadata associated with East Asia. It covers 110 UNESCO ICH elements associated with China, the Democratic People’s Republic of Korea, Japan, Mongolia, and the Republic of Korea.

The archive treats UNESCO concepts, country assignments, inscription years, list types, domains, and multinational nominations as structured institutional metadata. These fields describe how records are organized within the UNESCO information system and should not be interpreted as direct measures of cultural origin, cultural importance, prevalence, or national cultural diversity.

Dataset scope

The archive contains cleaned and derived data generated from a fixed, source-audited snapshot of publicly accessible UNESCO web resources.

The analytical dataset includes:

• 110 unique UNESCO ICH elements;
• 113 assignments to the five East Asian study countries;
• 105 distinct Primary Concepts;
• 273 distinct Secondary Concepts;
• element–country incidence tables;
• element–domain incidence tables;
• element–concept incidence tables;
• Primary and Secondary Concept co-occurrence networks;
• community assignments and country-community profiles;
• robustness, sensitivity, resampling, and leave-one-element-out outputs.

Reference numbers are used as the stable identifiers linking the derived tables.

Contents

The public archive includes:

• cleaned element-level metadata tables;
• country, domain, Primary Concept, and Secondary Concept incidence tables;
• network node and edge tables;
• Leiden and comparison-algorithm community assignments;
• country-level marginal occurrence and union-coverage outputs;
• diversity and country-similarity results;
• threshold and resolution sensitivity results;
• random-seed repeatability outputs;
• leave-one-element-out reconstruction results;
• pair-budget normalization and post-projection weighting diagnostics;
• cross-layer, multinational-nomination, and temporal sensitivity outputs;
• figure and supplementary-table source data;
• Python analysis scripts;
• configuration files and recorded random seeds;
• software-environment and package-version records;
• data dictionaries, file inventories, provenance records, and integrity hashes.

Reproducibility

Users should begin with:

• README.md
• RUN_REPRODUCTION.md
• DATA_DICTIONARY.csv or DATA_DICTIONARY.xlsx
• FILE_INVENTORY.csv
• MANIFEST.sha256

The archive is structured to reproduce the reported networks, community assignments, country profiles, robustness diagnostics, figures, and supplementary tables from the publicly released derived data.

Exact software and package versions are recorded in the environment files included in the archive.

Analytical features

The archive supports analysis of:

• institutional metadata structure;
• concept co-occurrence networks;
• bipartite-to-unipartite projection;
• Leiden community detection;
• comparison with Louvain and greedy modularity;
• random-seed repeatability;
• threshold and resolution sensitivity;
• leave-one-element-out stability;
• direct pair-budget normalization;
• post-projection cosine and association-strength weighting;
• non-exclusive country-community occurrence rates;
• exact recurrent-core union coverage;
• metadata diversity and country similarity;
• cross-layer concept associations;
• multinational nomination sensitivity;
• inscription-period sensitivity.

Applications

Potential applications include:

• reproducible cultural-heritage metadata research;
• auditing institutional classification and tagging systems;
• evaluating metadata quality and evidential limits;
• teaching network analysis and community detection;
• comparing projection and weighting strategies;
• developing robustness workflows for museum, archive, library, and heritage metadata;
• extending the analysis to later UNESCO metadata snapshots.

Usage notes

Country-community profiles are non-exclusive. One element may contain concepts belonging to several communities, so marginal occurrence rates may sum to more than 100%. These rates should not be interpreted as mutually exclusive national compositions.

The Primary and Secondary Concept networks are separate metadata layers and should not be combined without an explicitly defined cross-layer procedure.

Community results depend on edge-selection, projection, weighting, resolution, algorithm, and sample choices. The archive distinguishes random-seed repeatability, parameter sensitivity, algorithm agreement, sample perturbation, and snapshot stability rather than treating them as one general form of robustness.

Source and rights information

The released tables, analytical outputs, documentation, and code were generated by the authors.

Copies of complete third-party UNESCO exports, the UNESCO Dive graph source, and cached official webpages are not redistributed because reuse remains subject to the terms and rights indicated by the original sources.

The public archive instead provides source URLs, retrieval information, provenance metadata, and integrity hashes. A frozen source snapshot is retained by the authors and may be made available confidentially to editors or reviewers for verification, subject to applicable rights.

Funding: This work was supported by the Research Project of Hefei Philosophy and Social Sciences Planning (Grant No. HFSKYY202609); the Jiaxing Science and Technology Program Project (Public Welfare Research Special Project) (Grant No. 2025CGZ077); and the Scientific Research Start-up Project for Introducing Talents (PhD) of Anhui Jianzhu University (Grant No. 2023QDZ21).