How multilingual is Multilingual BERT?
This work addresses the problem of understanding and improving cross-lingual model transfer for NLP researchers and practitioners, though it is incremental as it analyzes an existing model.
The paper investigates the multilingual capabilities of Multilingual BERT (M-BERT), finding it effective for zero-shot cross-lingual transfer across 104 languages, with performance varying based on typological similarity and revealing systematic deficiencies in certain language pairs.
In this paper, we show that Multilingual BERT (M-BERT), released by Devlin et al. (2018) as a single language model pre-trained from monolingual corpora in 104 languages, is surprisingly good at zero-shot cross-lingual model transfer, in which task-specific annotations in one language are used to fine-tune the model for evaluation in another language. To understand why, we present a large number of probing experiments, showing that transfer is possible even to languages in different scripts, that transfer works best between typologically similar languages, that monolingual corpora can train models for code-switching, and that the model can find translation pairs. From these results, we can conclude that M-BERT does create multilingual representations, but that these representations exhibit systematic deficiencies affecting certain language pairs.