Multiscale Collaborative Deep Models for Neural Machine Translation
This addresses the training challenges for deep NMT models, offering incremental improvements in translation accuracy for machine translation systems.
The paper tackles the difficulty of training deep neural machine translation (NMT) models by proposing a MultiScale Collaborative (MSC) framework, which improves translation quality with gains of +2.2~+3.1 BLEU points on IWSLT tasks and a BLEU score of 30.56 on WMT14 English-German, outperforming state-of-the-art models.
Recent evidence reveals that Neural Machine Translation (NMT) models with deeper neural networks can be more effective but are difficult to train. In this paper, we present a MultiScale Collaborative (MSC) framework to ease the training of NMT models that are substantially deeper than those used previously. We explicitly boost the gradient back-propagation from top to bottom levels by introducing a block-scale collaboration mechanism into deep NMT models. Then, instead of forcing the whole encoder stack directly learns a desired representation, we let each encoder block learns a fine-grained representation and enhance it by encoding spatial dependencies using a context-scale collaboration. We provide empirical evidence showing that the MSC nets are easy to optimize and can obtain improvements of translation quality from considerably increased depth. On IWSLT translation tasks with three translation directions, our extremely deep models (with 72-layer encoders) surpass strong baselines by +2.2~+3.1 BLEU points. In addition, our deep MSC achieves a BLEU score of 30.56 on WMT14 English-German task that significantly outperforms state-of-the-art deep NMT models.