Sources and licences
Every word in the dictionary is attested in at least one of the sources below. They supply the words, spellings, conjugations and some example sentences; definitions and usage notes are written by language models following the method described separately.
DODa, Darija Open Dataset
The Moroccan lexicon, its conjugations, feminine and plural forms, Arabizi spellings and the real sentences quoted as examples come from DODa, published by Aissam Outchakoucht and Hamza Es-Samaali under the Creative Commons BY-NC 4.0 licence. Its data is used here with its authors’ permission. Translations of the quoted sentences are written by the model.
Algerian Darja corpus
Algerian words are ranked by their frequency in the Algerian-Darija corpus published by Ayoub Kirouane on Hugging Face under the Creative Commons BY 4.0 licence. The corpus is used to count words and spellings, to show the models how a word is used, and to feed the keyboard’s vocabulary; none of its texts is reproduced on the site.
Algerian-Darija on Hugging Face
awras Algerian dictionary
The word list published by awras under the Creative Commons BY 4.0 licence is used to confirm that an Algerian word exists and to collect its Latin-letter spellings. Its definitions and examples are not used: its own authors note that the current version contains errors.
algerian-darija-dictionary-v1 on Hugging Face
TArC, Tunisian Arabish Corpus
Tunisian lemmas, their Arabizi spellings and the real sentences quoted as examples come from TArC, published by Elisa Gugliotta and Marco Dinarelli under the Creative Commons BY-NC-SA 4.0 licence. Its data is used here with its authors’ permission. Quoted sentences are reproduced in the corpus’s Arabic-script transcription; their translations are written by the model.
Tunisian Derja corpus
Tunisian words are ranked by their frequency in the Tunisian Derja Dataset assembled by Linagora on Hugging Face, which brings together corpora published under their own licences. It is used to count words, to show the models how a word is used, and to feed the keyboard’s vocabulary; none of its texts is reproduced on the site.
Tunisian_Derja_Dataset on Hugging Face
Wiktionary
The French Wiktionary’s categories for Moroccan, Algerian and Tunisian Arabic are used to confirm that a word belongs to the dialect. Page titles give the word list; definitions are shown to the models only when they reread a word they did not recognise, and are not reproduced. Wiktionary content is published under the Creative Commons BY-SA 4.0 licence.
Language models
Definitions, usage notes, translations and written examples are by Claude Opus 5.5, checked against the independent opinion of Claude Sonnet 5.5 and judged by Claude Haiku 5.5, three Anthropic models. The English text is translated from the French by Claude Sonnet 5.5 and checked by Claude Haiku 5.5. None of it has been reviewed by a person.