Data and model licences
Mova's own code, data and models are under the Elamur Cosmic License (Elamur Cosmic License). It allows everything, except Big Brother.
Third-party data and anything derived from it keep the terms their sources require (section 7 of the license). A build gate (en::license::commercial_gate) refuses training sources whose license is unknown or restricts the purpose of use.
- Never shipped: data under non-commercial (NC), no-derivatives (ND) or research-only terms. Benchmarks under such terms are only measured against; their files are not in this repository.
- CC BY-SA, as little as possible: we quote it, test and measure on it, and learn from it. We do not ship it or modified copies of it. Files whose header declares CC BY-SA stay out of the free release (
tools/freesyncholds them). A model trained on CC BY-SA treebanks stays under CC BY-SA.
| path | source | license |
|---|---|---|
en/data/forms.tsv, en/data/lemmas.tsv | the AGID inflection database (Kevin Atkinson), plus word and tag frequencies from UD English EWT | AGID permissive license; the EWT-derived counts are CC BY-SA 4.0 |
en.bin (not in git; install.sh downloads it) | tokenizer, tagger, parser and lemmatizer trained by Mova on UD English EWT and ESLSpok plus the agreed school trees | CC BY-SA 4.0, as EWT |
gparse models (in the data directory, not in git) | Mova's GPU parser, same training data | CC BY-SA 4.0, as EWT |
en/data/train-licenses.tsv, en/data/ud-registry-en.tsv | the registry of training sources and their licenses | ours, Elamur Cosmic License |
en/seeds/**, global/seeds/** (except below), prag/data/cues.tsv, math/data/**, latex/data/**, mlab/data/** | written for Mova | ours, Elamur Cosmic License |
global/seeds/shortcuts/atomic-*.md | derived from ATOMIC-2020 (Hwang et al. 2021) | CC BY 4.0 |
| Core categories from Wikidata (generated, 2026-10-01) | derived from Wikidata | CC0 1.0 |
world/data/school/** | the school's courses and lessons, written for Mova by a large model as teacher | ours, Elamur Cosmic License |
world/data/** (fables) | questions on Aesop's fables (Project Gutenberg #21) | text public domain; questions ours, Elamur Cosmic License |
world/data/rules-induced.tsv | tree-repair rules induced by Mova (world induce) | ours, Elamur Cosmic License |
global/data/absurdity-roles.tsv, global/data/absurdity-roles-tale.tsv | the absurdity matrix: plausibility scores rated by a large model for Mova | ours, Elamur Cosmic License |
global/data/register-*.tsv | register lexicons: log-odds counts over public-domain Project Gutenberg books | ours, Elamur Cosmic License (texts public domain) |
tera/global/data/expressions.tsv, tera/global/data/idioms.tsv (closed part, not in the free release) | expressions and idioms with meanings, from the English Wiktionary and other open sources (STREUSLE, ConceptNet, MAGPIE, PARSEME 1.3, WordNet) | CC BY-SA 4.0 — internal, not in the free release |
mlab/tests/data/ext/nist/** | NIST StRD reference datasets | public domain (U.S. Government work) |
mlab/tests/data/ext/boost/** | Boost.Math test data | Boost Software License 1.0 (LICENSE_1_0.txt) |
*/tests/data/vmm/** | prompts and model outputs produced for Mova | ours, Elamur Cosmic License |
corpus/open/** (downloaded by train/get-corpus.sh) | see each folder's README | per folder |
docs/sefer-yetzirah/** | Sefer Yetzirah: Hebrew text (Warsaw 1884), English translations (Kalisch 1877, Westcott 1887/1893), commentaries, Mordell 1914 — each listed in its books.tsv | public domain; Sefaria's community translation CC0 |
** (not in git; made by tools/macula) | the Torah's dependency trees, converted from MACULA Hebrew Linguistic Datasets, available at github.com/Clear-Bible/macula-hebrew (Biblica, Clear Bible, the Groves Center, OpenScriptures); the Semantic Dictionary of Biblical Hebrew fields are not carried | CC BY 4.0 |
| ** (not in git) | the Torah in the JPS 1917 translation, from eBible.org | public domain |
tools/ocrfix/data/gt/**, tools/ocrfix/data/confusions.tsv | hand transcriptions of pages of public-domain printed editions (Cowley 1923, Schechter-Taylor 1899, Driver 1913, Ungnad 1911, Kautzsch 1884) and the confusable letters of the square script, made for Mova | the printed texts public domain; the transcriptions ours, Elamur Cosmic License |
** (not in git; made by tools/ocrfix) | OCR text of public-domain editions (Cowley 1923, Schechter-Taylor 1899) | public domain texts; the OCR ours |
papers/*.md | our summaries of papers and books; the papers keep their own licenses (named on each page) and are not copied here | the summaries are ours, Elamur Cosmic License |
This page in Markdown → https://md.mova.elamur.ai/free/en/docs/licence/DATA/