Cloze-driven Pretraining of Self-attention Networks

Alexei Baevski, Sergey Edunov, Yinhan Liu, Luke Zettlemoyer, Michael Auli

Abstract
We present a new approach for pretraining a bi-directional transformer model that provides significant performance gains across a variety of language understanding problems. Our model solves a cloze-style word reconstruction task, where each word is ablated and must be predicted given the rest of the text. Experiments demonstrate large performance gains on GLUE and new state of the art results on NER as well as constituency parsing benchmarks, consistent with BERT. We also present a detailed analysis of a number of factors that contribute to effective pretraining, including data domain and size, model capacity, and variations on the cloze objective.
Anthology ID:
D19-1539
Volume:
Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)
Month:
November
Year:
2019
Address:
Hong Kong, China
Editors:
Kentaro Inui, Jing Jiang, Vincent Ng, Xiaojun Wan
Venues:
EMNLP | IJCNLP
SIG:
SIGDAT
Publisher:
Association for Computational Linguistics
Note:
Pages:
5360–5369
Language:
URL:
https://aclanthology.org/D19-1539
DOI:
10.18653/v1/D19-1539
Bibkey:
Cite (ACL):
Alexei Baevski, Sergey Edunov, Yinhan Liu, Luke Zettlemoyer, and Michael Auli. 2019. Cloze-driven Pretraining of Self-attention Networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 5360–5369, Hong Kong, China. Association for Computational Linguistics.
Cite (Informal):
Cloze-driven Pretraining of Self-attention Networks (Baevski et al., EMNLP-IJCNLP 2019)
Copy Citation:
PDF:
https://preview.aclanthology.org/nschneid-patch-5/D19-1539.pdf
Data
CoNLL, CoNLL 2003, GLUE, MRPC, MultiNLI, Penn Treebank, QNLI, SST, SST-2
PDF Search