a b/biobert_ner/README.md 

  1
  
# BioBERT for NER





  
  

  2
  
To train an NER model with BioBERT-v1.1 (base), run the command below.





  
  

  3
  
<br>





  
  

  4
  
Before running this, make sure you have generated the pre-processed dataset using the generate_data.py file with the command mentioned in the parent directory. 





  
  

  5
  





  
  

  6
  
## Additional Requirements





  
  

  7
  
- seqeval: Used for NER evaluation (```pip install seqeval```)





  
  

  8
  





  
  

  9
  
## Training





  
  

  10
  
```





  
  

  11
  
export SAVE_DIR=./output





  
  

  12
  
export DATA_DIR=./dataset





  
  

  13
  





  
  

  14
  
export MAX_LENGTH=128





  
  

  15
  
export BATCH_SIZE=16





  
  

  16
  
export NUM_EPOCHS=5





  
  

  17
  
export SAVE_STEPS=1000





  
  

  18
  
export SEED=0





  
  

  19
  





  
  

  20
  
python run_ner.py \





  
  

  21
  
    --data_dir ${DATA_DIR}/ \





  
  

  22
  
    --labels ${DATA_DIR}/labels.txt \





  
  

  23
  
    --model_name_or_path dmis-lab/biobert-large-cased-v1.1 \





  
  

  24
  
    --output_dir ${SAVE_DIR}/ \





  
  

  25
  
    --max_seq_length ${MAX_LENGTH} \





  
  

  26
  
    --num_train_epochs ${NUM_EPOCHS} \





  
  

  27
  
    --per_device_train_batch_size ${BATCH_SIZE} \





  
  

  28
  
    --save_steps ${SAVE_STEPS} \





  
  

  29
  
    --seed ${SEED} \





  
  

  30
  
    --do_train \





  
  

  31
  
    --do_eval \





  
  

  32
  
    --do_predict \





  
  

  33
  
    --overwrite_output_dir





  
  

  34
  
```





  
  

  35
  





  
  

  36
  
## Results





  
  

  37
  
|             | precision |   recall | f1-score |





  
  

  38
  
|:---:|:---:|:---:|:---:|





  
  

  39
  
|         ADE |      0.6351 |     0.5680|      0.5997|





  
  

  40
  
|       Dosage |      0.9254  |    0.9485  |    0.9368   |





  
  

  41
  
|        Drug |      0.9580  |    0.9542  |    0.9561   |





  
  

  42
  
|         Duration |      0.8119  |    0.9021  |    0.8546   |





  
  

  43
  
|         Form |      0.9546  |    0.9456  |    0.9501   |





  
  

  44
  
|         Frequency |      0.9707  |    0.9668  |    0.9688   |





  
  

  45
  
|         Reason |      0.7203  |    0.7348  |    0.7275   |





  
  

  46
  
|         Route |      0.9530  |    0.9525  |    0.9527   |





  
  

  47
  
|         Strength |      0.9807  |    0.9846  |    0.9827   |





  
  

  48
  
|   micro avg |      0.9327  |    0.9330  |    0.9328   |





  
  

  49
  
|   macro avg |      0.9253  |    0.9225  |    0.9230   |