Facebook is pleased to announce the research award recipients for the Low-resource Neural Machine Translation (NMT) call for proposals. This effort is expected to contribute to the field of NMT through research into novel, strongly performing models under low-resource training conditions and/or comparable corpora mining techniques for low-resource language pairs.

Facebook selected the top 5 proposals. Of these, 3 were focused on low-resource modeling and 2 were focused on data mining approaches. The Principal Investigators are:

Trevor Cohn, University of Melbourne, Australia
Nearest neighbor search over vector space representations of massive corpora: An application to low-resource NMT

Victor O.K. Li, The University of Hong Kong, Hong Kong
Population-Based Meta-learning for Low-Resource Neural Machine Translation

David McAllester. Toyota Technological Institute at Chicago, USA
Phrase Based Unsupervised Machine Translation

Alexander Rush, Harvard University, USA
More Embeddings, Less Parameters: Unsupervised NMT by Learning to Reorder

William Wang, University of California, Santa Barbara, USA
Hierarchical Deep Reinforcement Learning for Semi-Supervised Low-Resource Comparable Corpora Mining