自回归的序列建模，两种序列属于不同的空间。

两篇文章：

1. Seq2Seq & Attention - Neural Machine Translation by Jointly Learning to Align and Translate
   1. Translate & Alignment
2. Seq2Seq & Local Attention - Effective Approaches to Attention-based Neural Machine Translation
   1. content-based function，考虑内容和位置(编码器和解码器)
   2. location-based function，只考虑位置(解码器)
   3. Monotonic、Predictive、Gaussian distribution

Neural Machine Translation by Jointly Learning to Align and Translate

机器翻译：machine translation

MNT：Neutral Machine Translation，神经网络机器翻译

Encoder - Decoder，Stop Token

SOTA：English 2 French Translation，fixed-length vector -> soft-alignment

align and translate

Translate，