A Sentence-Level Joint Relation Classification Model Based on Reinforcement Learning

Table 2

Experimental parameters.

Parameters	Values
Word dropout	0.5
Position embedding dim	5
Bi-LSTM dropout	0.75
Bi-LSTM hidden size	128
Word embedding dim	50
Batch size	128
Learning rate	0.02
Update rate	0.01
Regularization	0.0001
Sentence length	70
Sample times	3
Kernel size	60
Stride size	1
Sentence output size	230
Dropout	0.5