Neural Machine Translation in Linear Time

We present a novel neural network for processing sequences. The ByteNet is a one-dimensional convolutional neural network that is composed of two parts, one to encode the source sequence and the other to decode the target sequence. The two network parts are connected by stacking the decoder on top of the encoder and preserving the temporal resolution of the sequences. To address the differing lengths of the source and the target, we introduce an efficient mechanism by which the decoder is dynamically unfolded over the representation of the encoder. The ByteNet uses dilation in the convolutional layers to increase its receptive field. The resulting network has two core properties: it runs in time that is linear in the length of the sequences and it sidesteps the need for excessive memorization. The ByteNet decoder attains state-of-the-art performance on character-level language modelling and outperforms the previous best results obtained with recurrent networks. The ByteNet also achieves state-of-the-art performance on character-to-character machine translation on the English-to-German WMT translation task, surpassing comparable neural translation models that are based on recurrent networks with attentional pooling and run in quadratic time. We find that the latent alignment structure contained in the representations reflects the expected alignment between the tokens.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Language Modelling enwik8 ByteNet Bit per Character (BPC) 1.31 # 37
Machine Translation WMT2014 English-German ByteNet BLEU score 23.75 # 72
Hardware Burden None # 1
Operations per network pass None # 1
Machine Translation WMT2015 English-German ByteNet BLEU score 26.3 # 1

Methods


No methods listed for this paper. Add relevant methods here