Mova 0.5 — trained models (mova-model-0.5.tar.gz)

en.bin: the `en` model of Mova (tokenizer, tagger, dependency parser, morphology and lemmas), trained on the
Universal Dependencies English treebanks EWT and ESLSpok (CC BY-SA 4.0) and on Mova's own agreed school trees, with
the development and test sentences of every held-out set left out (2026-11-05).

gparse/: Mova's GPU parser model (gparse: a transformer encoder with biaffine arcs and relations), trained on the
same filtered data.

coref/gender.tsv: the coreference noun lexicon: gender, number and animacy counts of noun-pronoun dependency paths
over Mova's own en.bin parses of public-domain (EU/UA) Project Gutenberg tales and of Mova's school lessons. The
sources are public domain or Mova's own, but the trees come from a CC BY-SA model, so the table is under CC BY-SA 4.0.
It is a separate file; it holds no text of the books, only counts and sentence ids.

Licence: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0),
https://creativecommons.org/licenses/by-sa/4.0/

Attribution: Mova (ElamurAI); the Universal Dependencies treebanks UD_English-EWT and UD_English-ESLSpok
(https://universaldependencies.org). Project Gutenberg (https://www.gutenberg.org/) for the public-domain tales.
Parts of the training trees were patched by Anthropic's Claude models as teacher; Anthropic's terms restrict using
model outputs to develop competing models.

The code of Mova is under the Elamur Cosmic License (see the repository, https://github.com/ElamurAI/mova-free).