NLP-生成模型-2016-生成式摘要模型：Seq2Seq+Attention+Copy【Pointer网络的Copy机制解决Decoder端的OOV问题】【抽取式+生成式】【第一个生成式摘要模型】

柔光的暖阳◎ 2023-10-01 10:00 10阅读 0赞

[《原始论文：Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond》][Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond]

Seq2Seq（BiGRU-UniGRU）+Attention+Copy【Copy机制解决Decoder端的OOV问题】【第一个生成式摘要模型】【抽取式+生成式】

## 同一作者的抽取式摘要模型（效果比其生成式摘要模型好） ##

[《论文：SummaRuNNer: A Recurrent Neural Network based Sequence Model for Extractive Summarization of Documents》][SummaRuNNer_ A Recurrent Neural Network based Sequence Model for Extractive Summarization of Documents]

## 与PGNet的区别 ##

[《论文：Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond》][Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond]：不将Copy拷贝机制计算出的概率分布与在固定Vocab上计算的生成式的概率分布进行加和；（Copy机制只计算OOV词汇在源文本中所有词汇上的概率分布）

[《论文：Get To The Point: Summarization with Pointer-Generator Networks》][Get To The Point_ Summarization with Pointer-Generator Networks]：将Copy拷贝机制计算出的概率分布与在固定Vocab上计算的生成式的概率分布进行加和；

--------------------

参考资料：  
[【读】seq2seq—（3）Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond][seq2seq_3_Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond]

[Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond]: https://arxiv.org/abs/1602.06023
[SummaRuNNer_ A Recurrent Neural Network based Sequence Model for Extractive Summarization of Documents]: https://ui.adsabs.harvard.edu/abs/2016arXiv161104230N
[Get To The Point_ Summarization with Pointer-Generator Networks]: https://ui.adsabs.harvard.edu/abs/2018arXiv180511080C
[seq2seq_3_Abstractive Text Summarization using Sequence-to-sequence RNNs and Beyond]: https://blog.csdn.net/imsuhxz/article/details/82589606