rongyua / qajudge

QAJudge

Code and dataset of AAAI2020 Paper Iteratively Questioning and Answering for Interpretable Legal Judgment Prediction. paper

Fork from & See frame doc at https://github.com/haoxizhong/pytorch-worker.

Citation

Please cite our paper if you find it helpful.

@inproceedings{zhong2020iteratively,
    title={Iteratively Questioning and Answering for Interpretable Legal Judgment Prediction},
    author={Zhong, Haoxi and Wang, Yuzhong and Tu, Cunchao and Zhang, Tianyang and Liu, Zhiyuan and Sun, Maosong},
    booktitle = "Proceedings of AAAI",
    year = "2020"
}

Specific params for QAJudge

config/sample_qajudge.config is a sample config file for QAJudge.

Notice that some general parameters may be missing in this config.

[data]:

• train_formatter_type,valid_formatter_type,test_formatter_type: Use ZMDqn for crime prediction; Use FTDqn for article prediction.
• task: Use ft for article prediction (unrequired for others).

[model]:

• model_name: Use ZMDqn.

[rl]:

• batch_size: Required. Size of mini-batch.
• epsilone: Required. Param epsilon for greedy strategy.
• gamma: Required. Discount factor.
• target_update: Required. Period to update the target.
• memory_capacity: Required. Capacity of memory.
• n_actions: Required. Number of actions, that is, the size of question list.
• n_questions: Required. Chance to question, mentioned as K in the paper.

[ml]:

• lgb_path: The path of Predict Net model. You should put your Predict Net model named predict_net.pkl here.