Cwestiwn ysgrifenedig WQ100620
- Gofynnwyd gan
- John ClarkReform
- I'w ateb gan
- Cabinet Minister for Education and the Welsh Language
- Cyflwynwyd
- Dyddiad ateb
- Atebwyd
- Amser i'w ateb
- 13 diwrnod
Cwestiwn
Mae'r cwestiwn a'r ateb yn Saesneg gan nad oes fersiwn Gymraeg ar gael.
Will the AI model trained in partnership with Cosine be released under an open-source or open-weights license, or will it remain a closed, proprietary product?
Ateb
Atebwyd gan Cabinet Minister for Education and the Welsh Language (Anna Brychan)
The Welsh Government has not entered into a commercial agreement with Cosine. Following an approach from Cosine seeking advice on Welsh-language AI, the Welsh Government formed a partnership with Cosine based on signposting relevant open, publicly available resources. The Welsh Government's policy is to provide such advice and to signpost open resources applies to any organisation seeking advice in this area. This approach is consistent with the Welsh Government's policy of investing in the development, creation and collection of Welsh-language data, tools and resources through grant funding, with outputs released openly and publicly, and engaging with technology companies by signposting them to those open resources.
We would encourage more organisations to embed the Welsh language in their technologies. This is why the Welsh Government has invested, and continues to invest, in Welsh universities and open-source natural language processing research through its Welsh-language technology grant funding. Recent examples include Iriaith, NLP Cymru CyTag, Mapio Cymru and Wici Cymru a'i Phobl. All Welsh-language technology resources funded by the Welsh Government are made publicly available and free of charge for anyone to use and reuse.
The purpose of the partnership with Cosine is to help improve how AI works in Welsh and support the future of the language in the digital age. As the Welsh Government is only signposting publicly available resources, it does not determine how Cosine, or any other organisation, may reuse those resources, nor does it hold information on the licensing, release or commercial arrangements for any resulting model.
The Welsh Government suggested that Cosine include the following publicly available open resources:
• Transcriber, https://trawsgrifiwr.techiaith.cymru/
• Common Voice CY, https://commonvoice.mozilla.org/cy/datasets
• Paldaruo, https://techiaith.cymru/data/corpora/paldaruo/?lang=en
• Tagged sentences, https://github.com/techiaith/brawddegau-tagiedig
• DigiGrid, https://digigrid.cymru/
• CorCenCC, https://corpus.corcencc.org/?language=en
• Corpus of sentences https://techiaith.cymru/data/corpora/brawddegau/