PEARL

Official repository for PEARL: A Review-driven Persona-Knowledge Grounded Conversational Recommendation Dataset accepted at ACL Findings 2024.

Setup

1. Set environment

cd PEARL
conda create -n [YOUR_CONDA_ENV] python=3.10
conda activate [YOUR_CONDA_ENV]
pip install -r requirements.txt

2. Create an OpenAI API key here.

Dialogue Generation

1. Download resources.zip from gdrive.
2. Unzip resources.zip.
3. Extract vector_store_per_movie.tar.gz file.
4. mkdir -p data
5. Place vector_store_per_movie, abstracts_per_user.json, and dialogue_input_data.json under data.
6. Modify datagen/config/personal_info.json

• Replace [YOUR_API_KEY_NAME] and [YOUR_OPENAI_API_KEY] with your OpenAI API credentials.
• You may add more API keys to run the code more efficiently (make sure to set the API key name correctly in the script).

7. Modify scripts/generate_turn_by_turn.sh

• Replace [YOUR_API_KEY_NAME] with one of the keys from datagen/config/personal_info.json.

8. Run script

sh scripts/generate_turn_by_turn.sh