We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Microsoft Research Asia's Systems for WMT19

Abstract: 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).
Comments: Accepted to "Fourth Conference on Machine Translation (WMT19)"
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:1911.06191 [cs.CL]
  (or arXiv:1911.06191v1 [cs.CL] for this version)

Submission history

From: Yingce Xia [view email]
[v1] Thu, 7 Nov 2019 03:55:53 GMT (123kb,D)

Link back to: arXiv, form interface, contact.