Chapter recurrent modern/seq2seq (d2l-ai#865)

* fix errors in 9.7.0.

* fix errors 9.7.1. Encoder

* fix errors 9.7.2. Decoder

* fix errors in 9.7.3. Loss Function

* fix errors 9.7.4. Training

* fix errors 9.7.5. Prediction

* fix errors 9.7.6. Evaluation
fix errors 9.7.7. Summary
fix errors 9.7.8. Exercises

* fix invalid character in code

* change RNN to 循环神经网络

* replace GRU by 门控循环单元
replace LSTM by 长短期记忆网络

* fix errors in 9.7. Sequence to Sequence Learning

* fix errors in 9.7.1. Encoder

* "连结" easyly cause misunderstand.
"拼结" had more useage.

* fix errors in 9.7.2. Decoder

* fix errors in 9.7.3. Loss Function

* fix errors in concatenate's translation

* fix errors in 9.7.4. Training

* fix errors in 9.7.5. Prediction

* add term 'precision' and translation "精确度"

* fix errors in 9.7.6. Evaluation

* end  of fix errors in 9.7.

* fix translation errors

* Update seq2seq.md

* Update seq2seq.md

Co-authored-by: goldmermaid <[email protected]>

Author: zhuyuanxiang

TERMINOLOGY.md

@@ -114,6 +114,8 @@
 
 解码器，decoder
 
+精确度，precision
+
 经验风险最小化，empirical risk minimization
 
 局部最小值，local minimum
@@ -136,7 +138,7 @@
 
 拉普拉斯平滑，Laplace smoothing
 
-连结，concatenate
+拼接，concatenate
 
 类，class