STAGE: Span Tagging and Greedy Inference scheme for Aspect Sentiment Triplet Extraction

This repository contains Pytorch implementation for "STAGE: Span Tagging and Greedy Inference Scheme for Aspect Sentiment Triplet Extraction" (AAAI 2023) (AAAI Version and Arxiv Version)

1. Requirements

We conduct our experiments on Nvidia GeForce 3090 GPU, with CUDA version 11.6 and PyTorch v1.10.1.

To reproduce experimental environment.

conda create -n STAGE python=3.9
conda activate STAGE
python -m pip install -r requirements.txt

2. Data

We use ASTE-Data-V2-EMNLP2020 from https://github.com/xuuuluuu/SemEval-Triplet-data.git (widely-used datasets in ASTE task)

The data dir should be data/ASTE-Data-V2-EMNLP2020 (or , set the correct dataset_dir parameter during training or predicting)

3. Train

To reproduce our best test $F_1$ performance on four datasets:

python run.py

Best $F_1$ scores are shown in logs/best_score.txt when running on our environment.

We also provide our training log in logs/best_training.log. Please ignore the time information as another tasks were also running at the same time.

4. Evaluate

Change model_path, dataset, version variants in ``predict.py'' and run:

python predict.py

We provide the output file logs/best_16res_3D_predict.log

Citation

Please kindly cite our paper if this paper and the code are helpful.

@article{Liang2023stage,
   TITLE      = {STAGE: Span Tagging and Greedy Inference Scheme for Aspect Sentiment Triplet Extraction}, 
   VOLUME     = {37}, 
   URL        = {https://ojs.aaai.org/index.php/AAAI/article/view/26547}, 
   DOI        = {10.1609/aaai.v37i11.26547}, 
   NUMBER     = {11}, 
   JOURNAL    = {Proceedings of the AAAI Conference on Artificial Intelligence}, 
   AUTHOR     = {Liang, Shuo
               AND Wei, Wei
               AND Mao, Xian-ling
               AND Fu, Yuanyuan
               AND Fang, Rui
               AND Chen, Dangyang}, 
   YEAR       = {2023}, 
   MONTH      = {Jun.}, 
   PAGES      = {13174-13182} 
}