Open Relation Extraction in Patent Claims with a Hybrid Network

Table 3

Hyperparameter setting of the framework.

Layer	Hyperparameter
Embedding	Word embedding size	300
	POS embedding size	26
	Entity embedding size	12
CNN	Kernel size	3
	Number of filters	100
Bi-LSTM	State size	300
Dropout	Dropout rate	0.5
	Batch size	50
	Initial learning rate	0.01
	Sequence length	50