MDSBR (RecSys'25)

This is our Pytorch implementation for "MDSBR: Multimodal Denoising for Session-based Recommendation"

Abstract

Multimodal session-based recommendation (SBR) has emerged as a promising direction for capturing user intent using visual and textual item content. However, existing methods often overlook a fundamental issue: the modality features extracted from pre-trained models (e.g., BERT, CLIP) are inherently noisy and misaligned with user-specific preferences. This noise arises from label errors, task mismatch, and over-inclusion of irrelevant content, ultimately degrading recommendation quality. In this work, we propose a diffusion-based denoising framework that explicitly refines noisy pre-trained representations without full fine-tuning. By a structured denoising process, our Multimodal Denoising Diffusion Layer progressively eliminates the noise introduced by pre-trained models. Furthermore, we introduce two auxiliary modules: an Interest-Guided Denoising Layer that filters modality features using user interest, and a Multimodal Alignment Layer that enforces cross-modal coherence. Extensive experiments on real-world datasets demonstrate that our model significantly outperforms state-of-the-art methods while maintaining practical training efficiency.

Datasets

We use the Amazon datasets Cellphones, Grocery, Sports and Yelp. Make sure you download the datasets from the Amazon and Yelp.

Settings

python = 3.8
pytorch = 2.1.0
transformers = 4.36.2
cuda = 12.1 

DataProcessing

Enter the data folder for data processing:

cd preprocess

Then use the following files to process your data：

preprocess1.ipynb
preprocess2.ipynb
preprocess3.ipynb

Then run this command to get image and text about item:

python imagedownload.py

Then run this command to get image and text embedding:

python processimage.py
python processtext.py

Train

Please make sure all datas are in corresponding folder location, then run this command to Training and Prediction:

python main.py

Citation

If this work is useful for your research, please cite:

@inproceedings{MDSBR,
    author = {Li, Yutong and Zhang, Xinyi},
    title = {MDSBR: Multimodal Denoising for Session-based Recommendation},
    year = {2025},
    publisher = {Association for Computing Machinery},
    address = {New York, NY, USA},
    booktitle = {Proceedings of the Nineteenth ACM Conference on Recommender Systems},
    pages = {268–278},
    numpages = {11},
    series = {RecSys '25}
}