Awesome

HD-LoA Prompting

This repository contains the code for our paper:

"LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction" (ACL 2024)

Data

DocEE dataset can be download from this GitHub repository.

Please ensure data directory follows the structure below:

HD_LoA/
├── RAMS/
│   └── RAMS_1_0/
│       └── data/
│           ├── train.jsonlines
│           ├── dev.jsonlines
│           └── test.jsonlines
└── DocEE/
    └── data/
        ├── normal_setting/
        │   ├── train.json
        │   ├── dev.json
        │   └── test.json
        └── cross_domain_setting/
            ├── train_source_domain.json
            ├── train_targe_domain.json
            └── dev_targe_domain.json
            └── test_targe_domain.json

Run Experiment

To run HD-LoA prompting using the following command:

python main.py --dataset_name <dataset> --data_type <type> --model_type <model>

Acknowledgment

The code for accuracy evaluation is from PAIE. We appreciate their excellent contributions!

Citation

If you find our work useful, please consider citing:

@inproceedings{hdloaprompting2024,
  title={LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-Level Event Argument Extraction},
  author={Zhou, Hanzhang  and
    Qian, Junlang and
    Feng, Zijian and
    Lu, Hui and
    Zhu, Zixiao  and
    Mao, Kezhi},
  booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"},
  year={2024}
}