• Publications
  • Influence
Dual Learning for Machine Translation
TLDR
We develop a dual-learning mechanism, which can enable an NMT system to automatically learn from unlabeled data through a dual learning game. Expand
  • 478
  • 56
  • PDF
Deliberation Networks: Sequence Generation Beyond One-Pass Decoding
TLDR
This work was done when Yingce Xia, Lijun Wu and Jianxin Lin were interns at Microsoft Research. Expand
  • 119
  • 18
  • PDF
Incorporating BERT into Neural Machine Translation
TLDR
We propose a new algorithm named BERT-fused model, in which we first use BERT to extract representations for an input sequence, and then the representations are fused with each layer of the encoder and decoder of the NMT model through attention mechanisms. Expand
  • 56
  • 16
  • PDF
Adversarial Neural Machine Translation
TLDR
We study a new learning paradigm for Neural Machine Translation (NMT). Expand
  • 89
  • 14
  • PDF
Achieving Human Parity on Automatic Chinese to English News Translation
TLDR
We describe Microsoft's machine translation system and measure the quality of its translations on the widely used WMT 2017 news translation task from Chinese to English. Expand
  • 325
  • 12
  • PDF
Dual Supervised Learning
TLDR
We propose training the models of two dual tasks simultaneously, and explicitly exploiting the probabilistic correlation between them to regularize the training process. Expand
  • 75
  • 10
  • PDF
Depth Growing for Neural Machine Translation
TLDR
In this work, we propose an effective two-stage approach with three specially designed components to construct deeper NMT models, which result in significant improvements over the strong Transformer baselines on WMT$. Expand
  • 16
  • 8
  • PDF
Layer-Wise Coordination between Encoder and Decoder for Neural Machine Translation
TLDR
We propose layer-wise coordination for NMT, which explicitly coordinates the learning of hidden representations of the encoder and decoder together layer by layer, gradually from low level to high level. Expand
  • 64
  • 6
  • PDF
Conditional Image-to-Image Translation
TLDR
In this paper, we study a new problem, conditional image-to-image translation, which is to translate an image from the source domain to the target domain conditioned on a given image in the target. Expand
  • 65
  • 4
  • PDF
Budgeted Multi-Armed Bandits with Multiple Plays
TLDR
We study the multi-play budgeted multi-armed bandits (MP-BMAB) problem, in which pulling an arm receives both a random reward and a random cost, and a player pulls L(≥ 1) arms at each round. Expand
  • 47
  • 4
  • PDF