ChinsesNER-pytorch

train

setp 1: edit models/config.yml

embedding_size: 100
hidden_size: 128
model_path: models/
batch_size: 20
dropout: 0.5
tags:
  - ORG
  - PER

step 2: train

python3 main.py train
or
cn = ChineseNER("train")
cn.train()

...
epoch [4] |██████                   | 154/591
        loss 0.46
        evaluation
        ORG     recall 1.00     precision 1.00  f1 1.00
--------------------------------------------------
epoch [4] |██████                   | 155/591
        loss 1.47
        evaluation
        ORG     recall 0.92     precision 0.92  f1 0.92
--------------------------------------------------
epoch [4] |██████                   | 156/591
        loss 0.46
        evaluation
        ORG     recall 0.94     precision 1.00  f1 0.97

predict

python3 main.py predict
or 
cn = ChineseNER("predict")
cn.predict()

请输入文本: 海利装饰材料有限公司
[{'start': 0, 'stop': 10, 'word': '海利装饰材料有限公司', 'type': 'ORG'}]

REFERENCES

• Log-Linear Models, MEMMs, and CRFs
• Neural Architectures for Named Entity Recognition