🪸 Empowering Retrieval-based Conversational Recommendation with Contrasting User Preferences (NAACL'25)

This is the official repository for the NAACL 2025 paper: Empowering Retrieval-based Conversational Recommendation with Contrasting User Preferences

🧠 Overview

🪸CORAL is a retrieval-based CRS framework that explicitly represents and models the user, item, and contrasting preferences.

For more details, please refer to our 📄paper, 🛝slide, and 🌐blog(Korean).

📦 Installation

git clone https://github.com/kookeej/CORAL.git
cd CORAL
conda create -n coral python=3.10
conda activate coral
pip install -r requirements.txt

🗃️ Dataset

The datasets used in this work are available on Hugging Face.

Please download and extract the datasets into the data/ folder as follows:

data/
├── pearl/
├── inspired/
└── redial/

🚀 Training

Run the training scripts for each dataset:

# Pearl
bash run_pearl.sh

# Inspired
bash run_inspired.sh

# Redial
bash run_redial.sh

🧾 Citation

If you find this work helpful, please cite our paper:

@inproceedings{kook-etal-2025-empowering,
    title = "Empowering Retrieval-based Conversational Recommendation with Contrasting User Preferences",
    author = "Kook, Heejin  and
      Kim, Junyoung  and
      Park, Seongmin  and
      Lee, Jongwuk",
    booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2025.naacl-long.392/",
    pages = "7692--7707",
    ISBN = "979-8-89176-189-6",
}