Meta-Learning for Low-Resource Neural Machine Translation

TL;DR

This paper introduces a meta-learning approach to low-resource neural machine translation, enabling effective translation models with minimal training data by leveraging multilingual high-resource language tasks.

Contribution

It extends the MAML algorithm to NMT, utilizing universal lexical representations for cross-lingual transfer in low-resource settings.

Findings

Outperforms transfer learning methods in low-resource NMT

Achieves high BLEU scores with limited training data

Demonstrates effectiveness across 18 European and 5 diverse languages

Abstract

In this paper, we propose to extend the recently introduced model-agnostic meta-learning algorithm (MAML) for low-resource neural machine translation (NMT). We frame low-resource translation as a meta-learning problem, and we learn to adapt to low-resource languages based on multilingual high-resource language tasks. We use the universal lexical representation~\citep{gu2018universal} to overcome the input-output mismatch across different languages. We evaluate the proposed meta-learning strategy using eighteen European languages (Bg, Cs, Da, De, El, Es, Et, Fr, Hu, It, Lt, Nl, Pl, Pt, Sk, Sl, Sv and Ru) as source tasks and five diverse languages (Ro, Lv, Fi, Tr and Ko) as target tasks. We show that the proposed approach significantly outperforms the multilingual, transfer learning based approach~\citep{zoph2016transfer} and enables us to train a competitive NMT system with only a fraction of training examples. For instance, the proposed approach can achieve as high as 22.04 BLEU on Romanian-English WMT'16 by seeing only 16,000 translated words (~600 parallel sentences).