This is the BERT Named Entity Recognition code implements with PyTorch

• Download bert-based-cased.

• Run Steps:

  1. git clone git@github.com:unikcc/PyTorch-BERT-NER.git
  2. cd PyTorch-BERT-NER/scripts
  3. python preprocess.py
  4. sh run_ner.sh

• Result

       precision    recall  f1-score   support
  
   LOC    0.9659    0.9728    0.9694      1837
  MISC    0.9018    0.9067    0.9043       922
   PER    0.9714    0.9767    0.9740      1842
   ORG    0.9236    0.9292    0.9264      1341
   Weight   0.9481    0.9539    0.9510     5942 
   Micro-score   0.9481    0.9539    0.9510     5942 
   Macro-score   0.9407    0.9463    0.9435     5942