Modeling Activity-Driven Music Listening with PACE

Code for our paper "Modeling Activity-Driven Music Listening with PACE" by L. Marey, B. Sguerra and M. Moussallam, accepted for publication in the proceedings of 2024 ACM SIGIR Conference on Human Information Interaction and Retrieval (CHIIR'24).

Dependencies

pandas==2.1.2  
numba==0.58.1  
  
numpy==1.26.1  
scikit-learn==1.3.2  
statsmodels==0.14.0  
  
matplotlib==3.8.1  
seaborn==0.13.0  

Scripts

1. Raw user histories are trasformed into time series using process_raw_streams.py.
2. User answers to survey are prepared using process_raw_answers.py.
3. Dictionary Learning algorithm is run using compute_dictionary.py.
4. The selection of the best iteration in dictionary learning is done using choose_dictionary.py.
5. compute_baselines.py computes baselines scores and scores of PACE embeddings.
6. analyse_models.py plots logistic regression coefficients and related statistical reports.
7. make_fig1.py saves the plot of Figure 1.

Data

Input data folder must be organized as follows :

pace/
│
└── data/  
  └── raw/  
    ├── streams/  
    │ ├── one_year_all_respondents000000000000.csv  
    │ ├── ...  
    │ └── one_year_all_respondents0000000000399.csv  
    ├── other/  
    │ └── user_favorites.csv  
    └── answers/  
      └── records.csv 

Where one_year_all_respondents.csv files are stream history csv files with columns : user_id, ts_listen, media_id, context_id, context_type, listening_time, context_4.

records.csv being Records survey answer csv files, with columns ResponseId, uid, Status, Progress, Duration (in seconds), Q_consent, B_contexts_deezer_1, B_contexts_deezer_2, B_contexts_deezer_4, B_contexts_deezer_5, B_contexts_deezer_10, B_contexts_deezer_12, E_birth_year, E_age_range, E_gender.

And user_favorites.csv having columns user_id, item_id, item_type.

About alphacsc

The alphacsc package enables dictionary learning with multivariate time series : We slightly modified the package from version 0.4.0 to be able to extract information during dictionary learning.

Dupré La Tour, T., Moreau, T., Jas, M., & Gramfort, A. (2018). Multivariate Convolutional Sparse Coding for Electromagnetic Brain Signals. Advances in Neural Information Processing Systems (NIPS).
Jas, M., Dupré La Tour, T., Şimşekli, U., & Gramfort, A. (2017). Learning the Morphology of Brain Signals Using Alpha-Stable Convolutional Sparse Coding. Advances in Neural Information Processing Systems (NIPS), pages 1099–1108.

Contact

research@deezer.com