USER

This is the implementation for paper:

"Recommendation with Causality enhanced Natural Language Explanations." In WWW 2023.

For the implementation of the base models, we refer to the open source NLG4RS.

Overview

In this paper, we propose the task of debiased explainable recommendation for the first time. For solving this task, we build a principled framework to jointly correct the item- and feature-level biases, and design fault tolerant IPS mechanism and latent confounder modeling strategy to improve this framework.

Requirements

• Python 3.8
• Pytorch >=1.10.1

Datasets

We use three real-world datasets, including TripAdvisor-HongKong, Amazon-Movie&TV and Yelp Challenge 2019. All the datasets are available at this link.

Usage

• Download the codes and datasets.
• Run

For example: Run run_nete_user.py

python run_nete_user.py --dataset [dataset_name] --lr [learning_rate]

Acknowledgement

Any scientific publications that use our codes and datasets should cite the following paper as the reference:

@inproceedings{zhang2023recommendation,
  title={Recommendation with Causality enhanced Natural Language Explanations},
  author={Zhang, Jingsen and Chen, Xu and Tang, Jiakai and Shao, Weiqi and Dai, Quanyu and Dong, Zhenhua and Zhang, Rui},
  booktitle={Proceedings of the ACM Web Conference 2023},
  pages={876--886},
  year={2023}
}

If you have any questions for our paper or codes, please send an email to zhangjingsen@ruc.edu.cn.