Skip to content

Latest commit

 

History

History
41 lines (30 loc) · 2.56 KB

README.md

File metadata and controls

41 lines (30 loc) · 2.56 KB
Raw

SentiFMRecSys

This repository includes the code and data associated to our UMAP '24 paper Integrating sentiment features in factorization machines: Experiments on music recommender systems. Dataset and pretrained models have been uploaded separately due to size constraints: dataset and pretrained models (if any of these links do not work, contact alejandro.bellogin@uam.es, since they may have expired).

Steps

  1. Set up a Conda environment with all the necessary libraries:

    • Create environment: conda env create -f environment.yml
    • Activate environment: conda activate lastfm_sentiment_umap_24
    • If CUDA (GPU) isn't available at first, consider rebooting the system

  2. Train and evaluate models with the preprocessed dataset in dataset/lastfm_recbole-dataset.pth:

    • python3 recbole_run.py --dataset lastfm_recbole --config_files config/generic.yaml -m [Model]
    • Add --save option to save the model in "saved/" as .pth file

  3. Train and evaluate models by preprocessing the dataset from scratch (takes about ~8GB RAM):

    • python3 recbole_run.py --dataset lastfm_recbole --config_files config/generic_unprocessed.yaml -m [Model]

  4. Evaluate saved models:

    • python3 recbole_run.py --dataset lastfm_recbole --config_files config/generic.yaml --evaluate_model saved/[Saved Model].pth
    • Add --evaluation_mode [full | uniN | popN] to specify evaluation method, default is uni100
    • Some pretrained models are included, trained with the following embeddings:
      • v (valence)
      • a (arousal)
      • d (dominance)
      • stsc (sentiment ratio)

Citation

If you use our source code, dataset, or experiments for your research or development, please cite the following paper:

@inproceedings{wang2024umap,
  title={Integrating sentiment features in factorization machines: Experiments on music recommender systems},
  author={Javier Wang and Alejandro Bellogín and Iván Cantador},
  booktitle={{UMAP} '24: 32nd {ACM} Conference on User Modeling, Adaptation and Personalization, Cagliari, Italy, July 1 - 4, 2024},
  year={2024}
}