Microsoft Research Asia's Systems for WMT19
This work addresses machine translation challenges for news domains, demonstrating incremental improvements through integration of recent techniques.
The researchers tackled the WMT19 news translation tasks across 11 language directions, achieving first place in 8 and second place in 3, using Transformer-based systems enhanced with techniques like multi-agent dual learning and masked sequence-to-sequence pre-training.
We Microsoft Research Asia made submissions to 11 language directions in the WMT19 news translation tasks. We won the first place for 8 of the 11 directions and the second place for the other three. Our basic systems are built on Transformer, back translation and knowledge distillation. We integrate several of our rececent techniques to enhance the baseline systems: multi-agent dual learning (MADL), masked sequence-to-sequence pre-training (MASS), neural architecture optimization (NAO), and soft contextual data augmentation (SCA).