Sources & Licenses
Last updated: .
Shabdah's curated 5,000-concept set and all six content types shown on every card (translations, example sentences, illustrations, phrases, related words, and collocations) are Shabdah's own AI-generated content, audited before shipping. A small number of open external sources are used only to look up words and data that fall outside that curated set — never to study or generate the 5,000 concepts themselves. Every one of those sources is listed below, linked directly to where it lives and to the licence it's used under.
Japanese ↔ English
- JMdict/EDICT (licence: Creative Commons Attribution-ShareAlike 4.0) — dictionary lookups for words outside the curated 5,000.
- Tatoeba (licence: Creative Commons Attribution-ShareAlike 4.0) — example sentences.
- Tanaka Corpus (licence: Creative Commons Attribution-ShareAlike 4.0) — example sentences.
French ↔ English
- WOLF — Wordnet Libre du Français (licence: CeCILL-C) — French dictionary fallback for words outside the curated 5,000, mapped onto Princeton WordNet below.
- Princeton WordNet (licence: WordNet licence) — the English sense definitions behind the WOLF mapping above.
Hindi ↔ English
- FreeDict's eng-hin dictionary (licence: GPL-2) — Hindi dictionary fallback for words outside the curated 5,000. FreeDict's eng-hin is itself built from Shabdanjali, a dictionary released by IIIT Hyderabad's Language Technologies Research Centre and compiled by volunteer contributors, whose publishers make no claim of uniform quality — used only as a fallback, never as curated teaching content.
Shared across all three pairs
- FLORES-101 (Meta AI; licence: Creative Commons Attribution-ShareAlike 4.0) — parallel sentences used as a fallback across Japanese, Hindi, French and English.