Loading…
Three layers, three positions. Di heritage material free and e go remain free; di software no be free.
Di corpus text free to individual pipo, forever.
No paywall on di heritage material. No account required to read. We dey ask institutions to pay for institutional access; individual pipo, never.
Everything wey di project dey charge for dey outside dat line: software, licensing, convenience features, teaching, and commissioned work. Di date wey we formally publish dis commitment, and wey we no go shift am from, go enter here once di entity wey dey make am dey exist.
Prose, structured data and software get separate licences, because na three different kinds of work with three different reuse cases.
| Layer | Licence | Effect |
|---|---|---|
| Corpus prose | CC BY-SA 4.0 | Free to read, reuse and adapt with attribution; derivatives go remain open. |
| Structured data | CC BY 4.0 | E dey free for research and NLP use, plus African language tech work. |
| Software platform | Proprietary | We license am separately, and na di project main commercial line. |
Na intention, dem never grant am. Licence dey granted when dem publish am with di work, and di corpus never dey published under dis terms. Di choices above na di project settled intent and dem match di practice of comparable corpora; dem go become binding when di files ship carrying dem.
Dis two dey stand beside di licences as statement of intent, no be as licence terms.
Ifá verse content no dey licensed for generative AI training
Di downstream harm wey person fit already see pass any money wey e go bring, and di decision no even close.
Credit must still dey
Reuse wey strip di citations dey reproduce exactly di problem wey dis project dey exist to fix, since di citations na di substance, no be decoration.
Di pattern dey consistent: share-alike on prose, looser licence or outright waiver on structured data. Dis project dey follow am instead of inventing im own position.
Wikimedia
Wikidata structured claims and statements dey released under CC0. Wikipedia article prose na CC BY-SA.
Europeana
All metadata across Europeana na CC0. Prose inside di blogs and exhibitions na CC BY-SA, and di heritage media objects carry dia own individual rights statements.
Universal Dependencies
New treebank repositories dey default to CC BY-SA 4.0 for structured syntactic annotation, and providers free to use CC BY 4.0 instead.
CLARIN
CLARIN identify CC BY 4.0 and CC BY-SA 4.0 as di primary open licences for text corpora, covering both copyright and database rights, and e name CC0 as waiver tool instead of licence.
Academic use depend on citation wey resolve to specific passage inside specific version, and wey remain resolvable.
Ìpilẹ̀ṣẹ̀, "<document title>", https://ipilese.com/read/<section>/<slug>, we open am for <date>.
E never dey ground. No DOI dey, no versioned corpus release dey, and no snapshot dey deposited for any repository. Citation to dis work na URL plus access date for now, wey weak pass wetin scholarly resource suppose offer, and we name am as weaker instead of dressing am up.
Di practice wey we dey aim at na di one wey Perseus and DraCor establish: dereferenceable identifier wey resolve to passage inside named version, and versioned corpus snapshot wey dem deposit with DOI so dat computational analysis fit cite di exact release wey e run against.
One REST API na di source of truth for every surface, and na di same one wey anybody else for consume.
Documents, sections and dia metadata dey served over versioned JSON API. Responses carry di version wey dem serve am under, so client fit assert di version wey e really reach instead of di one wey e think say e request. Di endpoints wey we don implement start at ipilese.com/api/v1/documents