BLSTM-CRF-NER

reference:

paper:

Neural Architectures for Named Entity Recognition
End-toEnd Sequence labeling via BLSTM-CNN-CRF

code:

https://github.com/ZhixiuYe/NER-pytorch

requirement

python3.6

pytorch

Data Format

train data at ./dataset/aminer_train.dat

<word> <label>

usage:

train model:

python train.py

query:

1. run the server

   python server.py
   

2. GET Method:

   http://166.111.5.228:5011/query/<query>
   

3. Return in json, Example:

   http://166.111.5.228:5011/query/search some selection Thomas Edison State College 1902 Goel Shom 's papers
   {"LOC": ["selection", "1902"], "PER": ["Goel"], "CON": ["Shom's"], "DATE": [], "ORG": ["Edison", "State"], "KEY": ["Thomas", "College"], "O": ["search", "some", "papers"]}
   

File orgnization

|- train.py 
|- debug.py 
|- [dir] dataset (word library)
|- [dir] evaluation (help tools when training)
|- [dir] models (well-trained models)