"Data for DiMMRec"

"This dataset package supports the experiments of the paper \"Multi-Modal Disentanglement with Mutual-Information-Guided Mamba for Sequential Recommendation (DiMMRec)\". It is designed for research on multimodal sequential recommendation, where the goal is to predict a user's next interaction by modeling chronological user behavior together with multimodal item information.The package is constructed from three publicly available recommendation datasets, including MicroLens, Amazon-Baby, and Amazon-Games. These datasets cover short-video and e-commerce recommendation scenarios and provide different types of item-side multimodal information. For each dataset, duplicate user-item interactions are removed, and the remaining interactions are sorted chronologically to form user behavior sequences. The last two interactions of each user are held out for validation and testing, respectively, while all preceding interactions are used for training.The released files include processed user-item interaction sequences, item textual metadata, and visual resources. Specifically, sequence files are provided for sequential recommendation modeling; raw textual descriptions or item title lookup files are provided for text-based representation learning; and image URLs or packaged image files are provided for visual feature extraction."