YORO

Code for Paper "You Only Read Once: Constituency-Oriented Relational Graph Convolutional Network for Multi-Aspect Multi-Sentiment Classification"

AAAI2024

Yongqiang Zheng, Xia Li

Model

Requirements

• python==3.10.12
• torch==1.12.1+cu113
• transformers==4.30.2
• scikit-learn==1.2.2
• benepar==0.2.0

Datasets

Download datasets from these links and put them in the dataset folder:

• MAMS
• Rest14
• Lap14

Usage

1. Download Bing Liu's opinion lexicon

wget http://www.cs.uic.edu/\~liub/FBS/opinion-lexicon-English.rar
sudo apt-get install unrar
unrar x opinion-lexicon-English.rar
mv opinion-lexicon-English lexicon

3. Generate constituency-oriented graph

python graph.py

An example of the construction of Constituency-Oriented Relational Graph Convolutional Network (CorrGCN)

Training

bash train.sh

Credits

The code in this repository is based on SEGCN-ABSA.

Citation

@inproceedings{zheng2024you,
  title = {You Only Read Once: Constituency-Oriented Relational Graph Convolutional Network for Multi-Aspect Multi-Sentiment Classification},
  author = {Zheng, Yongqiang and Li, Xia},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  volume = {38},
  number = {17},
  pages = {19715--19723},
  year = {2024},
  url = {https://ojs.aaai.org/index.php/AAAI/article/view/29945},
  doi = {10.1609/aaai.v38i17.29945},
}