This repository contains code (will be available soon) and data for our paper DiaASQ: A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis
To clone the repository, please run the following command:
git clone https://github.com/unikcc/DiaASQ
In this work, we propose a new task named DiaASQ, which aims to extract Target-Aspect-Opinion-Sentiment quadruples from the given dialogue. You can find more details in our paper.
The model is implemented using PyTorch. The versions of the main packages used are shown below.
- python>=3.8
- attrdict>=2.0.1
- jieba>=0.42.1
- PyYAML>=6.0
- spacy>=3.4.2
- torch>=1.8.1
To set up the dependencies, you can run the following command:
pip install -r requirements.txt
You can download the source data from Google Drive Link
Then, unzip the files and place them under the data directory like the following:
./data/dataset/annotation_zh
./data/dataset/annotation_en
Generate JSON format files for Chinese data and English data(You should download ``)
python prepare_data.py --lang zh
python prepare_data.py --lang en
For example, the Chinese version train dataset with JSON format should locate at:
./data/dataset/json_zh/train.json
If you want to use our dataset, please cite the following paper:
@article{lietal2022arxiv,
title={DiaASQ: A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis},
author={Bobo Li, Hao Fei, Fei Li, Yuhan Wu, Jinsong Zhang, Shengqiong Wu, Jingye Li, Yijiang Liu, Lizi Liao, Tat-Seng Chua, Donghong Ji}
journal={arXiv preprint arXiv:2211.05705},
year={2022}
}