-
Code for the paper
Global Structure Knowledge-Guided Relation Extraction Method for Visually-Rich Document
. -
Congratulations! Our work has been accepted by the EMNLP2023 Findings conference.
We check the reproducibility under this environment.
- Python 3.8.18
- CUDA 11.1
To run the codes, you need to install the requirements:
git clone https://github.com/chenxn2020/GOSE.git
cd GOSE
conda create -n gose python=3.8
conda activate gose
pip install -r requirements.txt
We utilize LayoutXLM and LiLT as our backbone.
You can download the models and place them under the GOSE/
.
We provide example scripts for explaining the usage of our code. You can kindly run the following commands.
# Current path: */GOSE
bash standard.sh
# Current path: */GOSE
bash multi.sh
The repository benefits greatly from unilm/layoutlmft and LiLT. Thanks a lot for their excellent work.
If our paper helps your research, please cite it in your publication(s):
@article{DBLP:journals/corr/abs-2305-13850,
author = {Xiangnan Chen and
Juncheng Li and
Duo Dong and
Qian Xiao and
Jun Lin and
Xiaozhong Liu and
Siliang Tang},
title = {Global Structure Knowledge-Guided Relation Extraction Method for Visually-Rich
Document},
journal = {CoRR},
volume = {abs/2305.13850},
year = {2023},
url = {https://doi.org/10.48550/arXiv.2305.13850},
doi = {10.48550/ARXIV.2305.13850},
eprinttype = {arXiv},
eprint = {2305.13850},
timestamp = {Mon, 05 Jun 2023 15:42:15 +0200},
biburl = {https://dblp.org/rec/journals/corr/abs-2305-13850.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}