MT6: Multilingual Pretrained Text-to-Text Transformer with Translation Pairs

TL;DR

This paper introduces mT6, an enhanced multilingual text-to-text transformer that leverages translation pairs and novel training tasks to improve cross-lingual transfer across various NLP benchmarks.

Contribution

The paper proposes mT6, a multilingual pretraining method incorporating translation pairs and new objectives, advancing cross-lingual transfer capabilities beyond mT5.

Findings

mT6 outperforms mT5 on multiple benchmarks

Improved transferability across diverse NLP tasks

Effective use of translation pairs in pretraining

Abstract

Multilingual T5 (mT5) pretrains a sequence-to-sequence model on massive monolingual texts, which has shown promising results on many cross-lingual tasks. In this paper, we improve multilingual text-to-text transfer Transformer with translation pairs (mT6). Specifically, we explore three cross-lingual text-to-text pre-training tasks, namely, machine translation, translation pair span corruption, and translation span corruption. In addition, we propose a partially non-autoregressive objective for text-to-text pre-training. We evaluate the methods on eight multilingual benchmark datasets, including sentence classification, named entity recognition, question answering, and abstractive summarization. Experimental results show that the proposed mT6 improves cross-lingual transferability over mT5.