How to use this app

1. Start with the Overview. The table shows, field by field, how many records are complete and valid in each repository. The panels list the most frequent findings; click one to see the records it affects.
2. Read the strip. Every record carries ten cells, one per Dublin Core field: filled means present and valid, amber means present with a finding, an empty outline means missing. Hover over a strip to read it field by field.
3. Open Records. Search by title, author or DOI, or filter by repository, finding or score. Select a record to see each finding, why it matters, a proposed fix, the Schema.org JSON-LD and the original Dublin Core.
4. Go to Curation. Accept or reject the proposed fixes, for example canonical licence URIs or ISO dates. Export the accepted ones as JSON. The page also lists publisher spellings and possible duplicates.
5. Query the Knowledge graph. Click an example such as “Most frequent authors”, then change the SPARQL and run your own query. The engine runs locally in your browser.
6. Check About for the sources, the rules and the limits of this analysis.

Words used here

• OAI-PMH: the protocol repositories use to share their metadata with harvesters.
• Dublin Core (oai_dc): the simple, fifteen-field metadata format every OAI-PMH repository must offer.
• OpenAIRE guidelines: European conventions for writing access rights, publication types and identifiers in Dublin Core.
• Schema.org and DCAT: vocabularies for describing works on the web and catalogues of them.
• SPARQL: the query language for knowledge graphs.

Run it yourself

git clone https://github.com/sayamrina/metadata-lens.git && cd metadata-lens
python -m venv .venv && .venv/bin/pip install -r pipeline/requirements.txt
cd pipeline && ../.venv/bin/pytest -q
../.venv/bin/python -m metalens.pipeline --out ../web/public/data --since 2026-09-01 --limit 400
cd ../web && npm ci && npm run dev