DiaASQ

This repository contains data and code for the ACL23 (findings) paper: DiaASQ: A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis

See the project page for more details.

To clone the repository, please run the following command:

git clone https://github.com/unikcc/DiaASQ

News 🎉

:loudspeaker: 2023-05-10: Released code and dataset.
:zap: 2022-12-10: Created repository.

Overview

In this work, we propose a new task named DiaASQ, which aims to extract Target-Aspect-Opinion-Sentiment quadruples from the given dialogue. More details about the task can be found in our paper.

DiaASQ Data

The dataset can be found at:

  data/dataset
    - jsons_en
    - jsons_zh

Requirements

The model is implemented using PyTorch. The versions of the main packages:

python>=3.7
torch>=1.8.1

Install the other required packages:

pip install -r requirements.txt

Code Usage

Train && Evaluate on the Chinese dataset
```
bash scripts/train_zh.sh
```
Train && Evaluate on the English dataset
```
bash scripts/train_en.sh
```
GPU memory requirements

Dataset Batch size GPU Memory

Chinese 2 8GB.

English 2 16GB.
Customized hyperparameters:
You can set hyperparameters in main.py or src/config.yaml, and the former has a higher priority.

Dataset	Batch size	GPU Memory
Chinese	2	8GB.
English	2	16GB.

Citation

If you use our dataset, please cite the following paper:

@inproceedings{li-2023-diaasq,
    title = "{D}ia{ASQ}: A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis",
    author = "Li, Bobo  and Fei, Hao and Li, Fei and Wu, Yuhan and Zhang, Jinsong and Wu, Shengqiong and Li, Jingye and
      Liu, Yijiang and Liao, Lizi and Chua, Tat-Seng and Ji, Donghong",
    booktitle = "Findings of ACL",
    year = "2023",
    pages = "13449--13467",
}

unikcc / DiaASQ

readme