Skip to main content

Table 2 Hyperparameters used in our experiments

From: Family history information extraction via deep joint learning

Dimension of word embeddings50
Dimension of POS embeddings20
Dimension of label embeddings10
Number of LSTM hidden states100
Learning rate0.005
Dropout rate in entity recognition0.5
Dropout rate in relation extraction0.3
Epoch number20/25
Combination coefficient (α)0.4/0.5/0.6