Table 3 Detailed experimental settings
From: Deep learning with language models improves named entity recognition for PharmaCoNER
Parameters | Tune range | Optimal |
|---|---|---|
Sequence length | [128, 256, 300] | 300 |
Train batch size | [8, 16, 32] | 16 |
Dev batch size | 16 | 16 |
Test batch size | 16 | 16 |
Learning rate | [1e−05, 2e−05, 3e−05] | 2e−05 |
Epoch number | [10, 20, 30, 50] | 20 |
Warmup | 0.1 | 0.1 |
Dropout | 0.1 | 0.1 |