Chinese Nested Named Entity Recognition Based on Suffix Labeling

Our code mainly refers to W2NER. We optimize the span-based approach using the suffix labeling method,aiming to improve the performance of Chinese NER.

Environments

• python 3.8.19
• pytorch 2.4.0+cu118

Dependencies

• numpy (1.24.1)
• transformers (4.44.2)
• scikit-learn (1.3.2)
• prettytable (3.11.0)

Preparation

Download dataset
Process them to fit the same format as the example in data/
Put the processed data into the directory data/
Here is an example of the processed data

{"sentence": ["中", "国", "科", "学", "院", "大", "学"], "ner": [{"index": [0,1,2,3,4,5,6], "type": "organization"} ] }

We provide the code process_CMeEE.py for processing the original CMeEE dataset

Create configuration files(*.json) in configs/
We have provided configuration files about Weibo, Resume, OntoNotes4 and CMeEE-V2

Training

python main.py --config ./configs/example.json

The experiment records are stored in the directory train_logs/