Multi-task Hierarchical Cross-Attention Network for Multi-label Text Classification

Code for NLPCC2022 Task5 Track1 "Multi-label Classification Model for English Scientific Literature".

Multi-label Classification Model for English Scientific Literature: develop a multi-label classification model for scientific research literature based on the given metadata (title and abstract) of scientific research literature and corresponding hierarchical labels in a specific domain.

Requirements

python 3.8
transformers 3.4.0
pytorch 1.10.0+cu113
numpy 1.21.4
tqdm

Data

You can reconfigure settings in config.py.

Original dataset:

"train_data" : 'data/training set_modified.json'
"dev_data" : 'data/verification set_modified.json'

Provide a subset for testing:

"train_data" : 'data/test_training data.json'    #  (2000)
"dev_data" : 'data/test_verification data.json'  #  (500)

Running

python run.py --embedding [bert/glove] --model [NN/HAM] --start_train [True/False]

Results

The training results are saved in two files: [Modelname] _all_scores.txt [Modelname]_top5.txt

[Modelname] _all_scores.txt : The training set score and the validation set score for each epoch (including loss)

[Modelname]_top5_[scores_key].txt : The model's best five validation set scores (including loss)

Description:

Version: 1.1 Author: Lu Junyu, Zhang hao Date: 2022-04-19 15:26:12 LastEditors: Lu Junyu LastEditTime: 2022-07-29 15:26:12\

Cite

@inproceedings{lu2022multi,
  title={Multi-task Hierarchical Cross-Attention Network for Multi-label Text Classification},
  author={Lu, Junyu and Zhang, Hao and Shen, Zhexu and Shi, Kaiyuan and Yang, Liang and Xu, Bo and Zhang, Shaowu and Lin, Hongfei},
  booktitle={Natural Language Processing and Chinese Computing: 11th CCF International Conference, NLPCC 2022, Guilin, China, September 24--25, 2022, Proceedings, Part II},
  pages={156--167},
  year={2022},
  organization={Springer}
}