Skip to main content

Table 4 Results of the hyperparameter tuning, the last column shows the selected value for each language

From: Recent advances in Swedish and Spanish medical entity recognition in clinical texts using deep neural approaches

Hyperparameter

Evaluated values

Best Swedish/Spanish

Batch size

10, 20, 30, 40, 50, 100

30/10

Nr. of hidden units, LSTM

100, 200, 300, 400

300/300

Nr. of hidden units, char.

5, 50, 100, 150, 200

100/100

Learning rate

0.01, 0.005, 0.001, 0.0005, 0.00005

0.005/0.005

Drop-out

0.5, 0.8

0.5/0.5

Dimension character embeddings

50, 100, 150, 200, 300

100/300

Dimension word embeddings

100, 300

300/300