Parsing with Multilingual BERT, a Small Corpus, and a Small Treebank

TL;DR

This paper explores methods to adapt multilingual BERT for low-resource languages in dependency parsing, demonstrating significant improvements through language-specific pretraining and vocabulary augmentation.

Contribution

It introduces language-specific pretraining and vocabulary augmentation techniques to enhance multilingual BERT's performance on low-resource language parsing tasks.

Findings

Significant performance gains in low-resource dependency parsing

Effectiveness of language-specific pretraining and vocabulary augmentation

Importance of pretraining data relevance to target languages

Abstract

Pretrained multilingual contextual representations have shown great success, but due to the limits of their pretraining data, their benefits do not apply equally to all language varieties. This presents a challenge for language varieties unfamiliar to these models, whose labeled \emph{and unlabeled} data is too limited to train a monolingual model effectively. We propose the use of additional language-specific pretraining and vocabulary augmentation to adapt multilingual models to low-resource settings. Using dependency parsing of four diverse low-resource language varieties as a case study, we show that these methods significantly improve performance over baselines, especially in the lowest-resource cases, and demonstrate the importance of the relationship between such models' pretraining data and target language varieties.