Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data

This anonymized repository provides Python code to reproduce experiments from the paper "Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data", which has been accepted at the 18th ACM Conference on Recommender Systems (RecSys 2024) in the Reproducibility Track.

Abstract

This paper examines the influence of recommender systems on local music representation, discussing prior findings from an empirical study on the LFM-2b public dataset. This prior study argued that different recommender systems exhibit algorithmic biases shifting music consumption either towards or against local content. However, LFM-2b users do not reflect the diverse audience of music streaming services. To assess the robustness of this study's conclusions, we conduct a comparative analysis using proprietary listening data from a global music streaming service, which we publicly release alongside this paper. We observe significant differences in local music consumption patterns between our dataset and LFM-2b, suggesting that caution should be exercised when drawing conclusions on local music based solely on LFM-2b. Moreover, we show that the algorithmic biases exhibited in the original work vary in our dataset, and that several unexplored model parameters can significantly influence these biases and affect the study's conclusion on both datasets. Finally, we discuss the complexity of accurately labeling local music, emphasizing the risk of misleading conclusions due to unreliable, biased, or incomplete labels.

Datasets

Please download the datasets used in experiments in the following link: DEEZER DATASET

The dataset directory must follow this structure:

├── DEEZER_BR
│   └── DEEZER_BR.inter
├── DEEZER_FR
│   └── DEEZER_FR.inter
├── DEEZER_DE
│   └── DEEZER_DE.inter
├── DEEZER_GLOBAL
│   ├── user_country.csv
│   └── DEEZER_GLOBAL.inter
├── metadata_DEEZER_active.csv  
├── metadata_DEEZER_origin.csv 
└── metadata_DEEZER_musicbrainz.csv  

To create the folder architecture, run:

python create_folders.py

Environment

• python 3.10.12
• recbole 1.2.0
• pandas 2.2.1
• numpy 1.26.4
• seaborn 0.13.2
• matplotlib 3.8.4

Onboarding

We use the implementation of ItemKN and NeuMF available in RecBole, a Python library based on PyTorch that aims to provide a unified framework for developing and reproducing recommendation algorithms.

Change of the path of saved files must be made. Firstly, open the RecBole file trainer.py, using nano or any other text editor:

nano path_to_recbole/trainer/trainer.py

Add the following line in the begining of the file:

import pathlib

Then change the line:

saved_model_file = "{}-{}.pth".format(self.config["model"], get_local_time())

to:

saved_model_file = f"{self.config["platform"]}/{self.config["country"]}/{self.config["model"]}/get_local_time().pth"

Running the code

Training and testing a model on a specified dataset:

python run.py

Retrieve and save the Top K items for each user for each model:

python predict.py

Make figures

python figures.py