AMPERE: AMR-Aware Prefix for Generation-Based Event Argument Extraction Model

Code for our ACL-2023 paper AMPERE: AMR-Aware Prefix for Generation-Based Event Argument Extraction Model

Environment

• Python==3.7
• PyTorch==1.11.0
• ipdb==0.13.9
• transformers==4.18.0 (pip)
• tensorboardx==2.5.1 (pip)
• sentencepiece==0.1.96 (pip)
• penman==1.2.2 (pip)
• networkx==2.6.3 (pip)
• amrlib==0.7.1 (pip)

Or use the yml file we provide.

AMR Parser Setup

We follow the instruction in https://amrlib.readthedocs.io/en/latest/install/ to install the trained AMR parser model. We use parse_spring version 0.1.0 as our parser

Data

We support ace05e, and ere.

Preprocessing

Our preprocessing mainly adapts OneIE's and DEGREE's released scripts with minor modifications. We deeply thank the contribution from the authors of the paper.

ace05e

1. Prepare data processed from DyGIE++
2. Put the processed data into the folder processed_data/ace05e_dygieppformat
3. Run ./scripts/process_ace05e.sh

ere

1. Download ERE English data from LDC, specifically, "LDC2015E29_DEFT_Rich_ERE_English_Training_Annotation_V2", "LDC2015E68_DEFT_Rich_ERE_English_Training_Annotation_R2_V2", "LDC2015E78_DEFT_Rich_ERE_Chinese_and_English_Parallel_Annotation_V2"
2. Collect all these data under a directory with such setup:

ERE
├── LDC2015E29_DEFT_Rich_ERE_English_Training_Annotation_V2
│     ├── data
│     ├── docs
│     └── ...
├── LDC2015E68_DEFT_Rich_ERE_English_Training_Annotation_R2_V2
│     ├── data
│     ├── docs
│     └── ...
└── LDC2015E78_DEFT_Rich_ERE_Chinese_and_English_Parallel_Annotation_V2
      ├── data
      ├── docs
      └── ...

3. Run ./scripts/process_ere.sh

The above scripts will generate processed data in ./process_data.

Training

Run ./scripts/train_eae.sh

Clean up log when training

If you want to have more clean log file, you can comment out line 151 in the "layout.py" file in penman package

logger.info('Interpreted: %s', g)

This should be in CONDAENVPATH/envs/Ampere/lib/python3.7/site-packages/penman

Citation

If you find that the code is useful in your research, please consider citing our paper.

@inproceedings{acl2023ampere,
    author    = {I-Hung Hsu and Zhiyu Xie and Kuan-Hao Huang and Premkumar Natarajan and Nanyun Peng},
    title     = {AMPERE: AMR-Aware Prefix for Generation-Based Event Argument Extraction Model},
    booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL)},
    year      = {2023},
}

Contact

If you have any issue, please contact I-Hung Hsu at (ihunghsu@usc.edu)