Hierarchical information extraction Network (HiNet)

Source code for paper: HiNet: Novel Multi-Scenario & Multi-Task Learning with Hierarchical Information Extraction

Model architecture:

Bibtex:

@article{Zhou2023HiNet,
  title={HiNet: Novel Multi-Scenario & Multi-Task Learning with Hierarchical Information Extraction},
  author={Zhou, Jie and Cao, Xianshuai and Li, Wenhao and Bo, Lin and Luo, Chuan and Yu, Qian},
  year={2023},
}

Python >= 3.7  
Tensorflow >= 1.15.0  

python -u HiNet.py --config_file_path=../config/hinet_sample_schemas.json --task_type=train

You can modify the "FLAGS" parameters as needed.

• The work is supported by MeiTuan.
• The work is also supported by the National Natural Science Foundation of China (No. 62202025 and 62002012).

As of the latest time, HiNet model was deployed to several companies' business lines, and according to incomplete statistics, the business benefits of each company are as follows: