Metadata Mender

Reconcile and complete Zotero metadata from PubMed, OpenAlex, Crossref, Semantic Scholar, OpenAIRE, Unpaywall, CORE, and CRAN

Author

Raban Heller

What it is

Metadata Mender is a Zotero 9+ plugin that takes the items you already have and mends them: it looks up each item’s DOI, PMID, or CRAN package id against eight free databases and fills in the gaps — title, journal, volume/issue/pages, ISSN, authors, abstract, URL, publisher — while also recording the current citation count, an OpenAlex-based impact-factor surrogate, and the open-access PDF link when one exists.

It is not a search tool. It assumes you already curate your library in Zotero and simply want the metadata to be complete and correct.

Install

Compatibility: Zotero 9 and later.

Download the latest metadata-mender-<version>.xpi from the Releases page.
In Zotero open Tools → Plugins, click the gear icon, and choose Install Plugin From File….
Select the .xpi and restart Zotero if prompted.

Zotero plugin installer dialog — Installing the .xpi from Tools → Plugins.

Set up your contact email

Open Zotero → Settings → Metadata Mender and enter a contact email. This single field unlocks the polite-pool throughput at Crossref and is the required identifier for Unpaywall. Everything else (API keys for OpenAlex, NCBI, Semantic Scholar, CORE, Crossref Plus) is optional and only buys you more throughput.

Metadata Mender preferences pane — Preferences pane with contact email and optional API keys.

A typical run

Select one or more regular items in your library (notes and attachments are ignored).
Right-click and choose Mend metadata (DOI/PMID lookup).
A progress popup ticks per item. Press Esc at any time to stop after the current item.

For each item the plugin:

Reads the DOI (from the DOI field or the Extra field) and PMID (from Extra).
Queries the configured sources in priority order. The first source with a non-empty value for a given field wins.
If neither a DOI nor a PMID is present, optionally falls back to a Crossref title search and adopts the top match if its title is similar enough (Jaccard ≥ 0.7 on tokenised words). Toggle this in preferences.
Writes provenance, citation count, impact factor surrogate, and any open-access PDF link back to Extra.
Tags changed items with mended:YYYY-MM-DD (or a stable mended tag, per your tag policy).

Zotero right-click menu — Right-click context menu showing the “Mend metadata” action.

What gets written

Stable identifiers are appended once; temporal data (citations, impact factor, OA link) is rewritten on every run with today’s date so the series stays comparable across runs.

Line	Behaviour	Source
`PMID: <id>`	Append once	PubMed / Semantic Scholar
`PMCID: PMC<id>`	Append once	OpenAlex / PubMed / Semantic Scholar
`arXiv: <id>`	Append once	Semantic Scholar
`OpenAlex: W<id>`	Append once	OpenAlex
`Citations: <n> [<source>, YYYY-MM-DD]`	Rewritten every run	OpenAlex (preferred) → Crossref → Semantic Scholar
`Impact Factor: <x.xx> [OpenAlex 2yr mean citedness, YYYY-MM-DD]`	Rewritten every run	OpenAlex venue lookup
`OA-URL: <url>`	Rewritten every run	OpenAlex → Unpaywall → OpenAIRE → CORE
`Provenance: YYYY-MM-DD — field:source, …`	Rewritten on any change	derived from the merges actually performed

R packages

To mend a software item that represents an R package, give it either:

a URL of the form https://cran.r-project.org/package=ggplot2 in the Zotero URL field, or
a line CRAN: ggplot2 in Extra.

The plugin queries METACRAN (crandb.r-pkg.org) and fills in the package title, abstract, version, date, language (“R”), publisher (“CRAN”), authors (with [role] and ORCID annotations stripped), and URL, plus License:, Upstream-URL:, and CRAN: lines.

Reading the run summary

Done. N updated, N unchanged, N skipped, N not found, N errored.

updated — at least one field or Extra line changed; a mended:YYYY-MM-DD tag was added.
unchanged — sources answered but had nothing new.
skipped — item has no DOI/PMID and the title fallback is off or did not produce a confident match.
not found — every source returned 404.
errored — every source threw (network, 5xx after retries, etc.).

Per-item errors are surfaced inline (up to five) in the progress popup; full detail is always in Help → Debug Output.

Tips

Start with Fill empty fields only mode for your first run, then spot-check before switching to Overwrite existing fields.
The “impact factor” is OpenAlex’s 2-year mean citedness — a transparent surrogate, not the Clarivate JIF. It tracks closely for most journals but is not identical.
Author name parsing for PubMed’s “Surname IN” form is heuristic; spot-check authors after a bulk overwrite run.

Build from source

./build.sh

Produces metadata-mender-<version>.xpi. No toolchain required — the plugin is plain bootstrapped JS zipped into an XPI.

Cite

If you use Metadata Mender in published work, please cite it. The canonical metadata lives in CITATION.cff.

BibTeX

@software{heller_metadata_mender_2026,
  author  = {Heller, Raban},
  title   = {Metadata Mender: a Zotero plugin for reconciling item metadata
             against PubMed, OpenAlex, Crossref, Semantic Scholar, OpenAIRE,
             Unpaywall, CORE, and CRAN},
  year    = {2026},
  version = {0.6.0},
  license = {MIT},
  url     = {https://github.com/CTTIR/metadata-mender},
  note    = {R package / Zotero 9+ plugin}
}

APA

Heller, R. (2026). Metadata Mender: a Zotero plugin for reconciling item metadata against PubMed, OpenAlex, Crossref, Semantic Scholar, OpenAIRE, Unpaywall, CORE, and CRAN (Version 0.6.0) [Computer software]. https://github.com/CTTIR/metadata-mender