75 Languages, 1 Model: Parsing Universal Dependencies Universally

TL;DR

UDify is a multilingual model that leverages pretrained BERT to accurately predict various linguistic annotations across 75 languages, achieving state-of-the-art results without language-specific components.

Contribution

The paper introduces UDify, a novel multilingual multi-task model that fine-tunes a pretrained BERT on all UD datasets simultaneously, enabling high-quality parsing across many languages.

Findings

Low-resource languages benefit most from cross-linguistic training.

Multilingual training enables strong zero-shot parsing performance.

State-of-the-art scores achieved on multiple UD tasks.

Abstract

We present UDify, a multilingual multi-task model capable of accurately predicting universal part-of-speech, morphological features, lemmas, and dependency trees simultaneously for all 124 Universal Dependencies treebanks across 75 languages. By leveraging a multilingual BERT self-attention model pretrained on 104 languages, we found that fine-tuning it on all datasets concatenated together with simple softmax classifiers for each UD task can result in state-of-the-art UPOS, UFeats, Lemmas, UAS, and LAS scores, without requiring any recurrent or language-specific components. We evaluate UDify for multilingual learning, showing that low-resource languages benefit the most from cross-linguistic annotations. We also evaluate for zero-shot learning, with results suggesting that multilingual training provides strong UD predictions even for languages that neither UDify nor BERT have ever been trained on. Code for UDify is available at https://github.com/hyperparticle/udify.