.png){kind=icon}
Return
DReX: An explainable deep learning-based multimodal recommendation framework
.png){kind=icon}
PRE
AI
DOI：10.1016/j.knosys.2026.117184.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Multimodal recommender systems leverage diverse data sources, such as user interactions, content features, and contextual information, to address challenges like cold-start and data sparsity. However, existing methods often suffer from one or more key limitations, such as the isolated processing of different modalities, the requirement of complete multimodal data for each interaction during training, and the independent learning of user and item representations. These limitations can increase model complexity and hinder the learning of consistent user and item representations. To address these challenges, we propose DReX, a unified multimodal recommendation framework that maintains evolving global representations for both users and items. For each interaction, modality-specific features are combined into an interaction-level representation, which is selectively integrated to incrementally evolve the global user and item representations. This incremental update mechanism provides three key advantages: (1) simultaneous modeling of both nuanced interaction details and broader preference patterns, (2) eliminates the need for separate user and item feature extraction processes, leading to enhanced alignment in their learned representation, and (3) inherent robustness to varying or missing modalities. We evaluate the performance of the proposed approach on three real-world datasets containing reviews, ratings, and images as interaction modalities. By considering review text as a modality, our approach automatically generates interpretable keyword profiles for both users and items, which supplement the recommendation process with interpretable preference indicators. Experimental results demonstrate that our approach outperforms state-of-the-art methods across all evaluated datasets.
Keywords:
Multimodal
Explainability
Recommender system
Interaction-level features
Global features
Journal
K
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
Cited Papers
No cited papers available