Look Harder: A Neural Machine Translation Model with Hard Attention

39669

Knowledge Triples

Source Entity Relationship Target Entity
achieves
comparesWith
usedFor