Fully Non-autoregressive Neural Machine Translation: Tricks of the Trade
arXiv:2012.15833
Abstract
Fully non-autoregressive neural machine translation (NAT) is proposed to simultaneously predict tokens with single forward of neural networks, which significantly reduces the inference latency at the expense of quality drop compared to the Transformer baseline. In this work, we target on closing the performance gap while maintaining the latency advantage. We first inspect the fundamental issues of fully NAT models, and adopt dependency reduction in the learning space of output tokens as the basic guidance. Then, we revisit methods in four different aspects that have been proven effective for improving NAT models, and carefully combine these techniques with necessary modifications. Our extensive experiments on three translation benchmarks show that the proposed system achieves the new state-of-the-art results for fully NAT models, and obtains comparable performance with the autoregressive and iterative NAT systems. For instance, one of the proposed models achieves 27.49 BLEU points on WMT14 En-De with approximately 16.5X speed up at inference time.
9 pages
References in corpus (12)
- Distilling the Knowledge in a Neural Network
- fairseq: A Fast, Extensible Toolkit for Sequence Modeling
- High Fidelity Speech Synthesis with Adversarial Networks
- Aligned Cross Entropy for Non-Autoregressive Machine Translation
- KERMIT: Generative Insertion-Based Modeling for Sequences
- Semi-Autoregressive Training Improves Mask-Predict Decoding
- Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation
- Imitation Learning for Non-Autoregressive Neural Machine Translation
- Glancing Transformer for Non-Autoregressive Neural Machine Translation
- FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow
- Improving Fluency of Non-Autoregressive Machine Translation
- Learning to Recover from Multi-Modality Errors for Non-Autoregressive Neural Machine Translation