Datasets

The corpus data card, self-contained: it reads without opening any other page.

Identification

NameASKIP (African Sovereign Knowledge Infrastructure Program)
DOI10.5281/zenodo.21794559CONCEPT DOI: it designates the deposit and follows its versions. The version DOI — 10.5281/zenodo.21794560 for v1.1 — designates a snapshot and stops being the right one at the next release.
Referential versionv3.10.7-2026-08-06vocabulary the disease labels are normalised against
Aggregates materialised onAugust 9, 2026a DOI without this date points to a snapshot nothing places in time
Validated evidence63,227extracted from 72,932 — the check rejected 9,705
Hugging Faceaskip776/african-health-evidencesame corpus, tabular format
Observations49,511
Source publications9,1998,784 DOI
Countries documented42
Disease labels2,564
Researchers identified80
Languages53,235 EN · 19,676 FR
LicenceCC BY 4.0attribution required, commercial use permitted

The three limitations to know before using this data

They do not carry the same weight of proof, and their status is marked: COUNTED comes from a database count, READ from a published view, SUPPLIED from the extraction chain without recomputation here.

1. Nearly a third of observations have no countryREAD

15,472 / 49,511 (31 %)

Any geographic aggregation covers the rest. A country missing from this data is not a country without health problems: it is a country whose location could not be extracted. Read from public_api.coverage_reach.

2. Two thirds of observations have no usable dateCOUNTED

34,155 / 49,511 (69 %)

Their period is UNKNOWN, not old. No time series, no before/after comparison, no notion of “recent data” applies to this share. Obtained by difference between two read counts: total observations minus dated observations.

3. No human review, and the author chain is partialSUPPLIED

22.4% of the corpus traces back to an identified author

“Validated” means “passed an automatic check of form and plausibility”, never “verified by an expert”. An assertion can be well-formed, plausible, correctly quoted — and wrong. The 22.4% figure is supplied by the extraction chain and is not recomputed on this page; it is given for what it is.

Citing this corpus

ASKIP (African Sovereign Knowledge Infrastructure Program).
E-Shepha Hub. Referential v3.10.7-2026-08-06, aggregates of August 9, 2026.
63,227 validated evidence items of 72,932 extracted.
DOI 10.5281/zenodo.21794559. Licence CC BY 4.0. https://askip.e-shepha.com

The referential version and the aggregate date are part of the citation: without them, the reference points to a corpus state nothing allows you to recover.

Deposited on Zenodo under a permanent DOI, and on Hugging Face.