Abstractive Sentence Summarization with Attentive Recurrent Neural Networks 阅读笔记

简介:

Abstractive Sentence Summarization with Attentive Recurrent Neural Networks

  • Sumit Chopra et al., Facebook AI Research
  • NAACL2016
  • sentence level
  • encoder: 使用了基于注意力的CNN

    • 先将词的原始embedding(x_i)和位置embedding(l_i)(可训练)相加,作为词的full embedding(a_i)
    • 然后使用size=5的一维卷积核做一个卷积操作,得到aggregate embedding(z_i)
    • 计算attention:
      RAS_attention

    h_t-1是t-1时刻的隐层状态(吧)

    • 计算t时刻encoder的输出c_t:
      RAS_ct
  • decoder: 普通的RNN和LSTM都试了

    • 状态更新:
      RAS_decoder
  • 模型encoder的输入每次都是一个完整地句子,decoder每次要输出的时候,会将h_t-1给encoder,encoder根据句子和h_t-1计算attention生成c_t给decoder,然后decoder根据(y_t-1, h_t-1, c_t)计算要输出的单词。encoder还要更新position embedding(l_i)
  • 性能(RAS-Elman, k=10, k means beam size):

    • DUC-2004: Rouge-1:28.97/Rouge-2:8.26/Rouge-L:24.06
    • Gigaword: Rouge-1:33.78/Rouge-2:15.97/Rouge-L:31.15
目录
相关文章
|
机器学习/深度学习 算法
Keyphrase Extraction Using Deep Recurrent Neural Networks on Twitter论文解读
该论文针对Twitter网站的信息进行关键词提取,因为Twitter网站文章/对话长度受到限制,现有的方法通常效果会急剧下降。作者使用循环神经网络(recurrent neural network,RNN)来解决这一问题,相对于其他方法取得了更好的效果。
101 0
|
机器学习/深度学习 自然语言处理 算法
【文本分类】Convolutional Neural Networks for Sentence Classification
【文本分类】Convolutional Neural Networks for Sentence Classification
【文本分类】Convolutional Neural Networks for Sentence Classification
|
机器学习/深度学习 存储 人工智能
【文本分类】Recurrent Convolutional Neural Networks for Text Classification
【文本分类】Recurrent Convolutional Neural Networks for Text Classification
【文本分类】Recurrent Convolutional Neural Networks for Text Classification
|
机器学习/深度学习 大数据
【文本分类】Deep Pyramid Convolutional Neural Networks for Text Categorization
【文本分类】Deep Pyramid Convolutional Neural Networks for Text Categorization
144 0
【文本分类】Deep Pyramid Convolutional Neural Networks for Text Categorization
|
机器学习/深度学习 数据挖掘
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读
|
机器学习/深度学习 数据建模
2_Recurrent Neural Networks (RNNs)循环神经网络 —Simple RNNs
2_Recurrent Neural Networks (RNNs)循环神经网络 —Simple RNNs
189 0
2_Recurrent Neural Networks (RNNs)循环神经网络 —Simple RNNs
|
机器学习/深度学习 自然语言处理 前端开发
《Reasoning about Entailment with Neural Attention》阅读笔记
《Reasoning about Entailment with Neural Attention》阅读笔记
160 0
《Reasoning about Entailment with Neural Attention》阅读笔记
|
机器学习/深度学习 数据挖掘 计算机视觉
CV:翻译并解读2019《A Survey of the Recent Architectures of Deep Convolutional Neural Networks》第四章(一)
CV:翻译并解读2019《A Survey of the Recent Architectures of Deep Convolutional Neural Networks》第四章
CV:翻译并解读2019《A Survey of the Recent Architectures of Deep Convolutional Neural Networks》第四章(一)
|
机器学习/深度学习 存储 文字识别
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读(一)
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读
|
机器学习/深度学习 传感器 文字识别
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读(三)
Paper:《Generating Sequences With Recurrent Neural Networks》的翻译和解读
下一篇
无影云桌面