Task-Level Curriculum Learning for Non-Autoregressive Neural Machine Translation
arXiv:2007.08772
Abstract
Non-autoregressive translation (NAT) achieves faster inference speed but at the cost of worse accuracy compared with autoregressive translation (AT). Since AT and NAT can share model structure and AT is an easier task than NAT due to the explicit dependency on previous target-side tokens, a natural idea is to gradually shift the model training from the easier AT task to the harder NAT task. To smooth the shift from AT training to NAT training, in this paper, we introduce semi-autoregressive translation (SAT) as intermediate tasks. SAT contains a hyperparameter k, and each k value defines a SAT task with different degrees of parallelism. Specially, SAT covers AT and NAT as its special cases: it reduces to AT when k = 1 and to NAT when k = N (N is the length of target sentence). We design curriculum schedules to gradually shift k from 1 to N, with different pacing functions and number of tasks trained at the same time. We called our method as task-level curriculum learning for NAT (TCL-NAT). Experiments on IWSLT14 De-En, IWSLT16 En-De, WMT14 En-De and De-En datasets show that TCL-NAT achieves significant accuracy improvements over previous NAT baselines and reduces the performance gap between NAT and AT models to 1-2 BLEU points, demonstrating the effectiveness of our proposed method.
Accepted at IJCAI 2020 Main Track. Sole copyright holder is IJCAI
References in corpus (9)
- Sequence to Sequence Learning with Neural Networks
- Convolutional Sequence to Sequence Learning
- FastSpeech: Fast, Robust and Controllable Text to Speech
- MASS: Masked Sequence to Sequence Pre-training for Language Generation
- Levenshtein Transformer
- Sequence-Level Knowledge Distillation
- Hint-Based Training for Non-Autoregressive Machine Translation
- Semi-Autoregressive Neural Machine Translation
- Fine-Tuning by Curriculum Learning for Non-Autoregressive Neural Machine Translation