2020

Are All Languages Created Equal in Multilingual BERT?

Wu, Shijie, Dredze, Mark

Understand

Multilingual BERT (mBERT) trained on 104 languages has shown surprisingly good cross-lingual performance on several NLP tasks, even without explicit cross-lingual signals.

However, these evaluations have focused on cross-lingual transfer with high-resource languages, covering only a third of the languages covered by mBERT.
We explore how mBERT performs on a much wider set of languages, focusing on the quality of representation for low-resource languages, measured by within-language performance.
We consider three tasks: Named Entity Recognition (99 languages), Part-of-speech Tagging, and Dependency Parsing (54 languages each).

Built on

Roberta: A robustly optimized bert pretraining approach
Original
Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019 · 1907
Earlier work this paper cites.
On the cross-lingual transferability of monolingual representations
Original
Mikel Artetxe, Sebastian Ruder, and Dani Yogatama. 2019 · 1910
Earlier work this paper cites.
Unsupervised cross-lingual representation learning at scale
Original
Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2019 · 1911
Earlier work this paper cites.
Emerging cross-lingual structure in pretrained language models
Original
Shijie Wu, Alexis Conneau, Haoran Li, Luke Zettlemoyer, and Veselin Stoyanov. 2019 · 1911
Earlier work this paper cites.
Cross-lingual pre-training based transfer for zero-shot neural machine translation
Original
Baijun Ji, Zhirui Zhang, Xiangyu Duan, Min Zhang, Boxing Chen, and Weihua Luo. 2019 · 1912
Earlier work this paper cites.
Cross-lingual name tagging and linking for 282 languages
Xiaoman Pan, Boliang Zhang, Jonathan May, Joel Nothman, Kevin Knight, and Heng Ji. 2017 · 1958
Earlier work this paper cites.
Software Framework for Topic Modelling with Large Corpora
Radim Řehůřek and Petr Sojka. 2010 · 2010
Earlier work this paper cites.
Adam: A method for stochastic optimization
Original
Diederik P Kingma and Jimmy Ba. 2014 · 2014
Earlier work this paper cites.
Bridging nonlinearities and stochastic regularizers with gaussian error linear units
Dan Hendrycks and Kevin Gimpel. 2016 · 2016
Earlier work this paper cites.
Learning bilingual word embeddings with (almost) no bilingual data
Mikel Artetxe, Gorka Labaka, and Eneko Agirre. 2017 · 2017
Earlier work this paper cites.
Enriching word vectors with subword information
Piotr Bojanowski, Edouard Grave, Armand Joulin, and Tomas Mikolov. 2017 · 2017
Earlier work this paper cites.
Deep biaffine attention for neural dependency parsing
Timothy Dozat and Christopher D Manning. 2017 · 2017
Earlier work this paper cites.

Similar

Using the output embedding to improve language models
Ofir Press and Lior Wolf. 2017 · 2017
Cited alongside, same era.
Attention is all you need
Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017 · 2017
Cited alongside, same era.
Towards better UD parsing: Deep contextualized word embeddings, ensemble, and treebank concatenation
Wanxiang Che, Yijia Liu, Yuxuan Wang, Bo Zheng, and Ting Liu. 2018 · 2018
Cited alongside, same era.
XNLI: Evaluating cross-lingual sentence representations
Alexis Conneau, Ruty Rinott, Guillaume Lample, Adina Williams, Samuel Bowman, Holger Schwenk, and Veselin Stoyanov. 2018 · 2018
Cited alongside, same era.
Multilingual bert readme document
Jacob Devlin. 2018 · 2018
Cited alongside, same era.
BPEmb: Tokenization-free pre-trained subword embeddings in 275 languages
Benjamin Heinzerling and Michael Strube. 2018 · 2018
Cited alongside, same era.
Universal language model fine-tuning for text classification
Jeremy Howard and Sebastian Ruder. 2018 · 2018
Cited alongside, same era.
Subword regularization: Improving neural network translation models with multiple subword candidates
Taku Kudo. 2018 · 2018
Cited alongside, same era.
SentencePiece: A simple and language independent subword tokenizer and detokenizer for neural text processing
Taku Kudo and John Richardson. 2018 · 2018
Cited alongside, same era.
Word translation without parallel data
Guillaume Lample, Alexis Conneau, Marc’Aurelio Ranzato, Ludovic Denoyer, and Hervé Jégou. 2018 · 2018
Cited alongside, same era.
Universal dependencies 2.3
Joakim Nivre, Manying Zhang, and Hanzhi Zhu. 2018 · 2018
Cited alongside, same era.
Deep contextualized word representations
Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018 · 2018
Cited alongside, same era.

Then

Cloze-driven pretraining of self-attention networks
Alexei Baevski, Sergey Edunov, Yinhan Liu, Luke Zettlemoyer, and Michael Auli. 2019 · 2019
Later among the works it cites.
Cross-lingual language model pretraining
Alexis Conneau and Guillaume Lample. 2019 · 2019
Later among the works it cites.
BERT: Pre-training of deep bidirectional transformers for language understanding
Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019 · 2019
Later among the works it cites.
Sequence tagging with contextual and non-contextual subword representations: A multilingual evaluation
Benjamin Heinzerling and Michael Strube. 2019 · 2019
Later among the works it cites.
Unicoder: A universal language encoder by pre-training with multiple cross-lingual tasks
Haoyang Huang, Yaobo Liang, Nan Duan, Ming Gong, Linjun Shou, Daxin Jiang, and Ming Zhou. 2019 · 2019
Later among the works it cites.
Polyglot contextual representations improve crosslingual transfer
Phoebe Mulcaire, Jungo Kasai, and Noah A. Smith. 2019 · 2019
Later among the works it cites.
Pytorch: An imperative style, high-performance deep learning library
Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019 · 2019
Later among the works it cites.
How multilingual is multilingual BERT?
Telmo Pires, Eva Schlinger, and Dan Garrette. 2019 · 2019
Later among the works it cites.
Beto, bentz, becas: The surprising cross-lingual effectiveness of BERT
Shijie Wu and Mark Dredze. 2019 · 2019
Later among the works it cites.
Xlnet: Generalized autoregressive pretraining for language understanding
Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019 · 2019
Later among the works it cites.
Electra: Pre-training text encoders as discriminators rather than generators
Kevin Clark, Minh-Thang Luong, Quoc V. Le, and Christopher D. Manning. 2020 · 2020
Closest in time.
Cross-lingual ability of multilingual bert: An empirical study
Karthikeyan K, Zihan Wang, Stephen Mayhew, and Dan Roth. 2020 · 2020
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…