Resolution of English Event Pronouns Based on Machine Learning

Table 3

Experimental parameter configuration.

Bi-LSTM	Hidden dim: 200
	Number layers: 1
	Dropout: 0.2
FFNN	Hidden dim: 200
	Number layers: 1
	Dropout: 0.2
	Activation: RELU
Word Embedding	Embedding dim: 300
Character Embedding	Embedding dim: 100
Features	Pos	Embedding dim: 20
	Type
	Realis
	Genre
Optimizer	Adam
Learning rate	0.0001