Table 2:

CNN hyperparameters.

Hyperparameter	Value
vocabulary size	400,000
word embedding size	300
title max words	64
abstract max words	448
number of convolution filters	350
convolution filter sizes	2, 5, 8
dynamic max pooling number of regions	5
activation function for classification layer	sigmoid
activation function for all other layers	relu
hidden layer size	3365
journal embedding size	50
dropout rate	0.5
vocabulary dropout rate	0.25
batch size	128
learning rate	0.001