Frustratingly Simple Pretraining Alternatives to Masked Language Modeling

Atsuki Yamaguchi, George Chrysostomou, Katerina Margatina, Nikolaos Aletras


Abstract
Masked language modeling (MLM), a self-supervised pretraining objective, is widely used in natural language processing for learning text representations. MLM trains a model to predict a random sample of input tokens that have been replaced by a [MASK] placeholder in a multi-class setting over the entire vocabulary. When pretraining, it is common to use alongside MLM other auxiliary objectives on the token or sequence level to improve downstream performance (e.g. next sentence prediction). However, no previous work so far has attempted in examining whether other simpler linguistically intuitive or not objectives can be used standalone as main pretraining objectives. In this paper, we explore five simple pretraining objectives based on token-level classification tasks as replacements of MLM. Empirical results on GLUE and SQUAD show that our proposed methods achieve comparable or better performance to MLM using a BERT-BASE architecture. We further validate our methods using smaller models, showing that pretraining a model with 41% of the BERT-BASE’s parameters, BERT-MEDIUM results in only a 1% drop in GLUE scores with our best objective.
Anthology ID:
2021.emnlp-main.249
Volume:
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2021
Address:
Online and Punta Cana, Dominican Republic
Editors:
Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3116–3125
Language:
URL:
https://aclanthology.org/2021.emnlp-main.249
DOI:
10.18653/v1/2021.emnlp-main.249
Bibkey:
Cite (ACL):
Atsuki Yamaguchi, George Chrysostomou, Katerina Margatina, and Nikolaos Aletras. 2021. Frustratingly Simple Pretraining Alternatives to Masked Language Modeling. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3116–3125, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Frustratingly Simple Pretraining Alternatives to Masked Language Modeling (Yamaguchi et al., EMNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.emnlp-main.249.pdf
Video:
 https://aclanthology.org/2021.emnlp-main.249.mp4
Code
 gucci-j/light-transformer-emnlp2021
Data
GLUESQuAD