Are All Languages Created Equal in Multilingual BERT?
Preprint
Abstract
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). mBERT does better than or comparable to baselines on high resource languages but does much worse for low resource languages. Furthermore, monolingual BERT models for these languages do even worse. Paired with similar languages, the performance gap between monolingual BERT and mBERT can be narrowed. We find that better models for low resource languages require more efficient pretraining techniques or more data.Keywords
All Related Versions
This publication has 20 references indexed in Scilit:
- Cross-Lingual Pre-Training Based Transfer for Zero-Shot Neural Machine TranslationProceedings of the AAAI Conference on Artificial Intelligence, 2020
- Emerging Cross-lingual Structure in Pretrained Language Models2020
- Unsupervised Cross-lingual Representation Learning at Scale2020
- On the Cross-lingual Transferability of Monolingual Representations2020
- Unicoder: A Universal Language Encoder by Pre-training with Multiple Cross-lingual Tasks2019
- Beto, Bentz, Becas: The Surprising Cross-Lingual Effectiveness of BERT2019
- Cloze-driven Pretraining of Self-attention Networks2019
- XNLI: Evaluating Cross-lingual Sentence Representations2018
- SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing2018
- Enriching Word Vectors with Subword InformationTransactions of the Association for Computational Linguistics, 2017