Table 1 The layers used in the traditional network
From: Chemlistem: chemical named entity recognition using recurrent neural networks
Layer | Type | Input(s) | No. of output neurons | Notes |
|---|---|---|---|---|
te1 | Embedding | ti1 | 300 | |
tc1 | Conv1D | ti2 | 256 | Width = 3, activation = relu, dropout of 0.5 |
tm1 | Concatenate | te1, tc1 | 556 | |
tb1 | Bidirectional LSTM | tm1 | 64 per direction, total 128 | Dropout of 0.5 |
td1 | TimeDistributed Dense | tb1 | 5 | Activation = softmax |