Towards Bidirectional Hierarchical Representations for Attention-Based Neural Machine Translation | Read Paper on Bytez