OpenAlex Pathology Metadata
Open bibliographic metadata from OpenAlex covering the pathology literature — the only major dataset in this repo that contains no patient data at all.
What it is
Publication records retrieved from OpenAlex: authors, affiliations, references, and concept tags. The bibliometrics work uses open sources exclusively and explicitly no private or licensed datasets, which makes everything built on it fully shareable and reproducible.
Two properties shape how it must be used, and both are documented as pitfalls:
- It updates continuously. A query run today and tomorrow returns different results. The documented rule is to cache the metadata pull per analysis run; without a cached snapshot, no published figure can be regenerated.
- Author identity is inferred, not given. Disambiguation is never perfect — name variants, shared names, and transliteration all cause errors, the last being directly relevant for Turkish-language authorship. The documented rule is to report the ambiguity rate alongside any author-level metric.
The corpus boundary is a choice rather than a property of the data: restricting by journal, by concept, or by citation distance produces different networks and different “central” authors. How the group bounds it is not recorded. [unverified]
Why it matters for my work
It backs Bibliometrics — Who Cites Who and the WHO-CNS strand — asking who writes the WHO classification that the field then adopts. Because it needs no patient data, no ethics approval, and no de-identification, it is the group’s most frictionless dataset: the whole pipeline from query to published figure can live in the open, which is unusual in this repo.
How it connects
Citation Network Analysis — the method page for building and analysing the networks.
Bibliometrics — Who Cites Who — the project consuming it, with the published companion site.
De-identification — notable by its absence; this is the one dataset here where that whole layer of governance simply does not apply.
Open questions
- How is the corpus bounded for the pathology and WHO-CNS analyses? This drives every centrality result. [unverified]
- What is the measured disambiguation ambiguity rate? Required by the group’s own rule but not recorded. [unverified]
- Are cached snapshots archived alongside published figures so results can be regenerated exactly?