CPATT

Prompt-based Event Relation Identification with Constrained Prefix ATTention Mechanism

We propose a prompt-based method to identify the causal and temporal relation and employ a Constrained Prefix ATTention (CPATT) mechanism to ameliorate the prompt-tuning process.

Introduction of Code

We provide two different sets of codes for different datasets and experimental settings.

Each set of codes corresponds to a different experimental setup, but there are no structural changes.

Taking the Event Causality Identification as an example:

makedataset_eventstoryline.py corresponds to the template creation module.

train.py corresponds to the prompt-tunning module.

inference.py corresponds to the relation inference module.

Introduction of Data

Three different datasets are provided in the data folder.

Publication status

Zhang H, Ke W, Zhang J, et al. Prompt-based event relation identification with Constrained Prefix ATTention mechanism[J]. Knowledge-Based Systems, 2023: 111072.

@article{ZHANG2023111072,
title = {Prompt-based event relation identification with Constrained Prefix ATTention mechanism},
journal = {Knowledge-Based Systems},
volume = {281},
pages = {111072},
year = {2023},
issn = {0950-7051},
doi = {https://doi.org/10.1016/j.knosys.2023.111072},
url = {https://www.sciencedirect.com/science/article/pii/S0950705123008225},
author = {Hang Zhang and Wenjun Ke and Jianwei Zhang and Zhizhao Luo and Hewen Ma and Zhen Luan and Peng Wang},
}