Dual Learning for Machine Translation
2016·Arxiv