On-the-fly Cross-lingual Masking for Multilingual Pre-training

Xi Ai, Bin Fang


Abstract
In multilingual pre-training with the objective of MLM (masked language modeling) on multiple monolingual corpora, multilingual models only learn cross-linguality implicitly from isomorphic spaces formed by overlapping different language spaces due to the lack of explicit cross-lingual forward pass. In this work, we present CLPM (Cross-lingual Prototype Masking), a dynamic and token-wise masking scheme, for multilingual pre-training, using a special token [𝒞]x to replace a random token x in the input sentence. [𝒞]x is a cross-lingual prototype for x and then forms an explicit cross-lingual forward pass. We instantiate CLPM for the multilingual pre-training phase of UNMT (unsupervised neural machine translation), and experiments show that CLPM can consistently improve the performance of UNMT models on {De, Ro, Ne } ↔ En. Beyond UNMT or bilingual tasks, we show that CLPM can consistently improve the performance of multilingual models on cross-lingual classification.
Anthology ID:
2023.acl-long.49
Volume:
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
July
Year:
2023
Address:
Toronto, Canada
Editors:
Anna Rogers, Jordan Boyd-Graber, Naoaki Okazaki
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
855–876
Language:
URL:
https://aclanthology.org/2023.acl-long.49
DOI:
10.18653/v1/2023.acl-long.49
Bibkey:
Cite (ACL):
Xi Ai and Bin Fang. 2023. On-the-fly Cross-lingual Masking for Multilingual Pre-training. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 855–876, Toronto, Canada. Association for Computational Linguistics.
Cite (Informal):
On-the-fly Cross-lingual Masking for Multilingual Pre-training (Ai & Fang, ACL 2023)
Copy Citation:
PDF:
https://preview.aclanthology.org/dois-2013-emnlp/2023.acl-long.49.pdf
Video:
 https://preview.aclanthology.org/dois-2013-emnlp/2023.acl-long.49.mp4