https://github.com/ProHiryu/bert-chinese-ner

使用预训练语言模型BERT做中文NER
https://github.com/ProHiryu/bert-chinese-ner

bert chinese fine-tune ner tensorflow

Last synced: about 1 month ago
JSON representation

使用预训练语言模型BERT做中文NER

• Host: GitHub
• URL: https://github.com/ProHiryu/bert-chinese-ner
• Owner: ProHiryu
• License: mit
• Created: 2018-11-28T02:18:54.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2020-02-26T05:58:03.000Z (almost 5 years ago)
• Last Synced: 2024-08-02T08:09:59.074Z (4 months ago)
• Topics: bert, chinese, fine-tune, ner, tensorflow
• Language: Python
• Homepage:
• Size: 3.64 MB
• Stars: 924
• Watchers: 26
• Forks: 276
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-bert - ProHiryu/bert-chinese-ner
• awesome-transformer-nlp - ProHiryu/bert-chinese-ner - Use the pre-trained language model BERT to do Chinese NER. (Tasks / Named-Entity Recognition (NER))

README

        
# bert-chinese-ner

## 前言

使用预训练语言模型BERT做中文NER尝试，fine - tune BERT模型

PS: 移步最新[**albert fine-tune ner**](https://github.com/ProHiryu/albert-chinese-ner)模型

## 代码参考

- [BERT-NER](https://github.com/kyzhouhzau/BERT-NER)

- [BERT-TF](https://github.com/google-research/bert)

## 使用方法

从[BERT-TF](https://github.com/google-research/bert)下载bert源代码，存放在路径下bert文件夹中

从[BERT-Base Chinese](https://storage.googleapis.com/bert_models/2018_11_03/chinese_L-12_H-768_A-12.zip)下载模型，存放在checkpoint文件夹下

使用BIO数据标注模式，使用人民日报经典数据

train：

`python BERT_NER.py --data_dir=data/ --bert_config_file=checkpoint/bert_config.json --init_checkpoint=checkpoint/bert_model.ckpt --vocab_file=vocab.txt --output_dir=./output/result_dir/`

## 结果

经过100个epoch跑出来的结果

```

eval_f = 0.9662649

eval_precision = 0.9668882

eval_recall = 0.9656949

global_step = 135181

loss = 40.160034

```

测试结果第一句：

![](test.png)