Table 1 System hyperparameters for each PharmaCoNER run
Parameter | Run 1 | Run 2 | Run 3 | Run 4 |
|---|---|---|---|---|
Sense-disambiguation embedding dimension | 128 | 128 | 128 | 128 |
Pre-trained word embeddings | FastText-SBC + Reddit | W2V-SBWC + Reddit | FastText-SBWC + Reddit | SNOMED-SBC + Reddit |
Word embeddings dimension | 300 | 300 | 300 | 300 |
Character embedding dimension | 50 | 50 | 50 | 50 |
Hidden layers dimension (for each LSTM) | 100 | 100 | 100 | 100 |
Learning method | SGD | SGD | SGD | SGD |
Dropout rate | 0.5 | 0.5 | 0.5 | 0.5 |
Learning rate | 0.005 | 0.005 | 0.005 | 0.005 |
Epochs | 100 | 100 | 100 | 100 |