Reinforcement Learning based Curriculum Optimization for Neural Machine Translation

TL;DR

This paper introduces a reinforcement learning framework to automatically optimize training curricula for neural machine translation, improving translation quality by effectively utilizing heterogeneous data during training.

Contribution

It proposes a novel reinforcement learning-based method to learn optimal curricula jointly with NMT, outperforming baseline approaches and matching state-of-the-art manual curricula.

Findings

Achieved up to +3.4 BLEU improvement over baselines.

Outperformed uniform and filtering data strategies.

Matched performance of hand-designed curricula.

Abstract

We consider the problem of making efficient use of heterogeneous training data in neural machine translation (NMT). Specifically, given a training dataset with a sentence-level feature such as noise, we seek an optimal curriculum, or order for presenting examples to the system during training. Our curriculum framework allows examples to appear an arbitrary number of times, and thus generalizes data weighting, filtering, and fine-tuning schemes. Rather than relying on prior knowledge to design a curriculum, we use reinforcement learning to learn one automatically, jointly with the NMT system, in the course of a single training run. We show that this approach can beat uniform and filtering baselines on Paracrawl and WMT English-to-French datasets by up to +3.4 BLEU, and match the performance of a hand-designed, state-of-the-art curriculum.