nmT5 - Is parallel data still relevant for pre-training massively multilingual language models?

Mihir Kale, Aditya Siddhant, Rami Al-Rfou, Linting Xue, Noah Constant, Melvin Johnson

Browse the full ACL paper archive.