Reinforcement Learning based Curriculum Optimization for Neural Machine Translation
arXiv:1903.00041
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.
NAACL 2019 short paper. Reviewer comments not yet addressed
References in corpus (5)
- Lingvo: a Modular and Scalable Framework for Sequence-to-Sequence Modeling
- Fast Domain Adaptation for Neural Machine Translation
- Curriculum Learning and Minibatch Bucketing in Neural Machine Translation
- An Empirical Exploration of Curriculum Learning for Neural Machine Translation
- Neural Machine Translation Training in a Multi-Domain Scenario