主要用途:
- certain segments in the input sequence are selectively replicated in the output sequence.
- COPYNET can nicely integrate the regular way of word generation in the decoder with the new copy-ing mechanism which can choose sub-sequences in the input sequence and put them at proper places in the output se-quence.
- 即:可以有选择性的从源端复制句子到合适的目标端。
3.copynet
3.1. model overreview
网络架构图
Encoder:
- 和别的encoder一样。
Decoder端:
read M and predict the target sequence,和之前的decoder有以下几点不同:
- Prediction:COPYNET基于两种混合模式的概率模型来预测单词。包括生成模式和copy模式,copy模式是直接从源端pick单词到目标端。
- State_update:t-1时刻的单词被用来更新t时刻的状态,但是copynet不仅用t-1时刻单词的word_embeddi