Hi All,
I am happy to announce that we have added four new language pairs into the WMT22 General MT task:
Ukrainian <-> Czech
Ukrainian <-> English
Croatian <-> English
Livonian <-> English
And we are working to collect resources for Yakut<->Russian. Testsets of these language pairs are likely not going to contain multiple domains. The domain for Ukrainian>CS/EN is going to be focused on humanitarian needs.
If you know about any available resources not listed on the WMT webpage for Ukrainian<>Czech/English, let us know.
We would be happy to welcome any additional sponsor who could support human evaluation with Ukrainian bilingual annotators.
We introduce a new approach to download training data simply via command line: https://www.statmt.org/wmt22/mtdata/index.html
Have a lovely day,
Tom
(in Germany, he/him)