Improved Multilingual Language Model Pretraining for Social Media Text via Translation Pair Prediction

TL;DR

This paper introduces translation pair prediction as a pretraining task to enhance zero-shot multilingual transfer of mBERT on social media tasks, showing significant improvements especially for challenging language pairs.

Contribution

The study proposes a novel pretraining task called translation pair prediction (TPP) to improve multilingual transfer in social media NLP tasks, especially for difficult language pairs.

Findings

37% average relative improvement in NER F1 across target languages

12% relative improvement in sentiment classification F1

6.7% accuracy boost in POS tagging

Abstract

We evaluate a simple approach to improving zero-shot multilingual transfer of mBERT on social media corpus by adding a pretraining task called translation pair prediction (TPP), which predicts whether a pair of cross-lingual texts are a valid translation. Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a model on source language task data and evaluate the model in the target language. In particular, we focus on language pairs where transfer learning is difficult for mBERT: those where source and target languages are different in script, vocabulary, and linguistic typology. We show improvements from TPP pretraining over mBERT alone in zero-shot transfer from English to Hindi, Arabic, and Japanese on two social media tasks: NER (a 37% average relative improvement in F1 across target languages) and sentiment classification (12% relative improvement in F1) on social media text, while also benchmarking on a non-social media task of Universal Dependency POS tagging (6.7% relative improvement in accuracy). Our results are promising given the lack of social media bitext corpus. Our code can be found at: https://github.com/twitter-research/multilingual-alignment-tpp.