ACSA-HGCN icon indicating copy to clipboard operation
ACSA-HGCN copied to clipboard

The code of COLING2020 paper "Aspect-Category based Sentiment Analysis with Hierarchical Graph Convolutional Network".

ACSA-HGCN

Aspect-Category based Sentiment Analysis with Hierarchical Graph Convolutional Network [COLING 2020].

Requirements

  • Python 3.7
  • Pytorch 1.5.1

Running

Modify the corresponding BERT_BASE_DIR, DATA_DIR and output_dir to run the script.

sh run.sh

Note

The statistical numbers of the Explicit and Implicit Aspects test set in the first paragraph of section 4.5 were put in the wrong order. The Explicit Aspects test set contains 360 samples and the Implicit Aspects test set contains 212 samples.

Citation

If the code is used in your research, please cite our paper as follows:

@inproceedings{cai2020aspect,
  title={Aspect-Category based Sentiment Analysis with Hierarchical Graph Convolutional Network},
  author={Cai, Hongjie and Tu, Yaofeng and Zhou, Xiangsheng and Yu, Jianfei and Xia, Rui},
  booktitle={Proceedings of the 28th International Conference on Computational Linguistics},
  pages={833--843},
  year={2020}
}