TY  - RPRT
TI  - nmT5 -- Is parallel data still relevant for pre-training massively multilingual language models?
AU  - Mihir Kale
AU  - Aditya Siddhant
AU  - Noah Constant
AU  - Melvin Johnson
AU  - Rami Al-Rfou
AU  - Linting Xue
PY  - 2021
UR  - https://arxiv.org/abs/2106.02171
ID  - 2106.02171
ER  -