Att-BLSTM

Model

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

模型主要包括五部分:

  1. Input layer
  2. Embedding layer
  3. Lstm layer
  4. Attention layer
  5. Output layer

Word Embedding

没啥好说的…

Bidirectional Network

Bi-LSTM结构, 最后输出

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

Attention

Attention部分, 先对LSTM输出做非线性**, 然后计算attention权重.

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

最终的句表示:

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

Classifying

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

损失函数: 加上L2正则的交叉熵

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

实验

数据集: SemEval-2010
参数设置:
word embedding: 分别尝试了50, 100
optimizer: AdaDelta
learning rate: 1.0
batch size: 10
dropout: 0.3, 0.3, 0.5
评价指标: macro-F1score

Attention-Based Bidirectional Long Short-Term Memory Networks for Relation Classification阅读笔记

相关文章:

  • 2021-04-03
  • 2021-10-29
  • 2021-07-21
  • 2022-01-20
  • 2021-11-03
  • 2021-04-23
  • 2021-09-28
  • 2021-07-30
猜你喜欢
  • 2022-01-24
  • 2021-06-09
  • 2021-09-03
  • 2021-05-08
  • 2021-08-15
  • 2021-07-23
  • 2021-04-16
相关资源
相似解决方案