Code for CONSEQUENCES'23 accepted paper titled, "A First Look at Selection Bias in Preference Elicitation for Recommendation".
To generate synthetic topics for items, we create a user-item graph and use graph embeddings, using BiNE method.
To run the BiNE network embedding, create an conda env using the "req_bine.txt" file provided with the repo.
$conda create -n --file req_bine.txt
For running the main code, create a conda env using the "req.txt" file provided with the repo.
$conda create -n --file req.txt
- download the Yahoo! R3 dataset.
- download the Coat dataset
Put them in the ./data/raw/yahoo, and ./data/raw/coat folders respectively.
The python file ./src/simulator.py has the code to generate the two fully-synthetic datasets (from coat and yahoo), and one fully-synthetic dataset.
$python src/simulator.py --data coat
The dataset is generated at line #73, #86, and #97 in the code. Following that the dataset can be used for different downstream tasks.
To run the script for different datasets, run the following script
$ sbatch run_$DATASET.sh