Mova corpus 0.5 (corpus-0.5.tar.gz)

Licence: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0), for the whole corpus, as for
Mova's trained models, https://creativecommons.org/licenses/by-sa/4.0/

Attribution: Mova corpus 0.5, the Elamur team (https://github.com/ElamurAI/mova-free). Each folder's README says
where its data comes from. The sources, cited:

books/
- Project Gutenberg, https://www.gutenberg.org/ (public domain in the USA and in the EU/UA): Lewis Carroll,
  Alice's Adventures in Wonderland (1865), ebook #11; Rudyard Kipling, Just So Stories (1902), ebook #2781; Kenneth
  Grahame, The Wind in the Willows (1908), ebook #289. Changes: the Project Gutenberg header and footer are removed.

treebanks/
- Annotation scheme: Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre and Daniel Zeman. 2021.
  Universal Dependencies. Computational Linguistics 47(2): 255-308. https://universaldependencies.org (UD 2.18, the
  conventions of UD English EWT: Natalia Silveira et al. 2014, A Gold Standard Dependency Corpus for English,
  LREC 2014).
- mova-school-agreed-en.conllu: the sentences of Mova's school lessons (CC BY-SA 4.0, the Elamur team). Drafted by
  Mova's parser (trained on UD English EWT and ESLSpok, CC BY-SA 4.0), patched by Anthropic's Claude models as
  teacher, and reconciled by Mova's written rules with three open witness parsers: Stanza (Peng Qi, Yuhao Zhang,
  Yuhui Zhang, Jason Bolton and Christopher D. Manning. 2020. Stanza: A Python Natural Language Processing Toolkit for
  Many Human Languages. ACL 2020 System Demonstrations; Apache-2.0), Trankit (Minh Van Nguyen, Viet Dac Lai, Amir Pouran
  Ben Veyseh and Thien Huu Nguyen. 2021. Trankit: A Light-Weight Transformer-based Toolkit for Multilingual Natural
  Language Processing. EACL 2021 System Demonstrations; Apache-2.0) and spaCy (Matthew Honnibal, Ines Montani, Sofie
  Van Landeghem and Adriane Boyd. 2020. spaCy: Industrial-strength Natural Language Processing in Python.
  doi:10.5281/zenodo.1212303; MIT).
- mova-silver-en.conllu: Tatoeba sentences under CC0 1.0 (https://tatoeba.org, each with its sentence id in
  `# source`), example sentences from Mova's seeds, and six sentences from papers of the ACL Anthology
  (https://aclanthology.org) under CC BY 4.0, each paper named in `# source`.
- mova-bronze-en.conllu: example sentences from Mova's seeds.

math-scripts/
- Arkil Patel, Satwik Bhattamishra and Navin Goyal. 2021. Are NLP Models really able to Solve Simple Math Word
  Problems? NAACL 2021. SVAMP, https://github.com/arkilpatel/SVAMP, MIT licence (the training split).
- Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry
  Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse and John Schulman. 2021. Training Verifiers to Solve Math
  Word Problems. arXiv:2110.14168. GSM8K, https://github.com/openai/grade-school-math, MIT licence (the training
  split). The scripts are Mova's, written by a large model and checked by Mova's arithmetic core.

contrast/
- Jason Weston, Antoine Bordes, Sumit Chopra, Alexander M. Rush, Bart van Merriënboer, Armand Joulin and Tomas
  Mikolov. 2015. Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks. arXiv:1502.05698. bAbI
  v1.2, CC BY 3.0. Changes: names, places, things and verbs replaced (`world babi-contrast`).
- StepGame phrasings: Mova's own, after Zhengxiang Shi, Qiang Zhang and Aldo Lipani. 2022. StepGame: A New Benchmark
  for Robust Multi-Hop Spatial Reasoning in Texts. AAAI 2022 (no StepGame text is included).
- Contrast sets: Matt Gardner et al. 2020. Evaluating Models' Local Decision Boundaries via Contrast Sets. Findings
  of EMNLP 2020.

Parts were written or patched by Anthropic's Claude models as teacher; Anthropic's terms restrict using model outputs
to develop competing models. No non-commercial, research-only or restricted text is in this corpus.

The code of Mova is under the Elamur Cosmic License, which covers only code
(https://github.com/ElamurAI/mova-free/blob/main/LICENSE.md); it does not cover this corpus.
