Table 4 Experimental parameters settings
From: A hybrid method based on semi-supervised learning for relation extraction in Chinese EMRs
Parameters | value |
|---|---|
Batch size | 64 |
Dimension of character embedding | 300 |
Dimension of position embedding | 25 |
GRU hidden units | 512 |
GRU hidden layer | 3 |
Window size | 3,5,7 |
Number of filters | 128 |
Learning rate | 0.015 |
Optimizer | Adam |
Dropout | 0.5 |