BERT-NER

Use google BERT to do CoNLL-2003 NER !

Try to implement NER work based on google's BERT code!

First git clone https://github.com/google-research/bert.git

Second download file in this project

Third download bert snapshot, extract and rename folder checkpoint

BERT
|____ bert
|____ BERT_NER.py
|____ checkpoint
|____ output

Third run:

  python BERT_NER.py   \
                  --task_name="NER"  \ 
                  --do_train=True   \
                  --do_eval=True   \
                  --do_predict=True
                  --data_dir=NERdata   \
                  --vocab_file=checkpoint/vocab.txt  \ 
                  --bert_config_file=checkpoint/bert_config.json \  
                  --init_checkpoint=checkpoint/bert_model.ckpt   \
                  --max_seq_length=128   \
                  --train_batch_size=32   \
                  --learning_rate=2e-5   \
                  --num_train_epochs=3.0   \
                  --output_dir=./output/result_dir/ 

result:

The predicted result is placed in folder ./output/result_dir/. It contain two files, token_test.txt is the tokens and label_test.txt is the labels for each token. If you want a more accurate evaluation result you can use script conlleval.pl for evaluation.

The following evaluation results differ from the evaluation results specified by conll2003.

注：For the parameters of the above model, I have not made any modifications. All parameters are based on the BERT default parameters. The better parameters for this problem can be adjusted by yourselves.

The f_score evaluation codes come from:https://github.com/guillaumegenthial/tf_metrics/blob/master/tf_metrics/__init__.py

reference:

• https://github.com/google-research/bert

• https://arxiv.org/abs/1810.04805