Open data
Data
What the site publishes in machine-readable form: the indicators of the active project, the publication feed, and the public corpus to come. Every field is documented; every value carries its reading date.
The indicator file
The file data/impact.json publishes the indicators of the active project, regenerated at each measurement campaign. Its address is stable, it reads without a key or an account, and it is served with headers that allow third-party programs to read it. One command is enough:
curl https://research.odersa.org/data/impact.jsonIts content at the reading, exactly as served:
{
"date": "2026-08-28",
"langues_traitees": 1000,
"fiches_de_faits": 3444,
"domaines": 8,
"langues_document_pays": 935,
"langues_document_espece": 932,
"langues_document_substance": 924,
"langues_document_figure": 132,
"valeurs_comparees_equivalence": 415,
"conflits_equivalence": 0,
"generations_repetees": 147,
"empreintes_identiques": 147,
"controles_automatiques": 218
}The field dictionary
A field that cannot be understood without asking us is not open. The table documents every field of the file: what the value measures, and how to read it.
| Field | Value | Reading |
|---|---|---|
date | 2026-08-28 | Date of the reading: a datum of the result, not an archive note. |
langues_traitees | 1000 | Languages processed by the system, equal treatment. |
fiches_de_faits | 3444 | Structured fact sheets held. |
domaines | 8 | Fact domains covered. |
langues_document_pays | 935 | Languages producing a complete document, country trial subject (Kenya). |
langues_document_espece | 932 | Languages producing a complete document, species trial subject (lion). |
langues_document_substance | 924 | Languages producing a complete document, substance trial subject. |
langues_document_figure | 132 | Languages producing a complete document, historical-figure trial subject – the least covered domain, published at the same rank. |
valeurs_comparees_equivalence | 415 | Numerical values compared across languages by the external verifier, trial subject, at the published threshold. |
conflits_equivalence | 0 | Conflicts found by that comparison. |
generations_repetees | 147 | Generations repeated in two separate processes. |
empreintes_identiques | 147 | Identical SHA-256 fingerprints across the two processes. |
controles_automatiques | 218 | Automated checks accompanying the system’s development. |
The data contract
A published value is dated, regenerable, documented
The file’s address is stable. Existing fields keep their meaning from one reading to the next; the dictionary may grow, never contradict itself, and any change is announced on the News page. A field this page does not explain has no place in the file.
The site’s other data
- The publication feed – The dated facts of the News page, in Atom format: readable by any aggregator, with no sign-up and no account.
- The public corpus, in preparation – Unretouched system outputs, under the CC BY 4.0 licence. Until it is released, the Demonstrations page shows dated outputs.
Every cited value carries its measurement date. The site’s citation format, BibTeX included, is given on the Publications page.