Joint Similarity Item Exploration and Overlapped User Guidance for Multi-Modal Cross-Domain Recommendation

TL;DR

This work tackles Multi-Modal Cross-Domain Recommendation (MMCDR) where items carry multi-modal signals and users overlap sparsely across domains. It introduces SIEOUG, a tripartite framework combining similarity item exploration, user-item collaborative filtering, and overlapped user guidance to simultaneously exploit intra-domain multimodal signals and cross-domain knowledge transfer. Key innovations include robust item similarity graph fusion, sparse group-wise item hypergraphs, Guidance-based Optimal User Matching via Wasserstein optimization, and a cross-domain contrastive objective that leverages overlapped users for positive transfer. Experiments on Amazon datasets demonstrate state-of-the-art performance across tasks and show robust gains even with low overlap, highlighting SIEOUG’s practical impact for sparse, multimodal cross-domain settings.

Abstract

Cross-Domain Recommendation (CDR) has been widely investigated for solving long-standing data sparsity problem via knowledge sharing across domains. In this paper, we focus on the Multi-Modal Cross-Domain Recommendation (MMCDR) problem where different items have multi-modal information while few users are overlapped across domains. MMCDR is particularly challenging in two aspects: fully exploiting diverse multi-modal information within each domain and leveraging useful knowledge transfer across domains. However, previous methods fail to cluster items with similar characteristics while filtering out inherit noises within different modalities, hurdling the model performance. What is worse, conventional CDR models primarily rely on overlapped users for domain adaptation, making them ill-equipped to handle scenarios where the majority of users are non-overlapped. To fill this gap, we propose Joint Similarity Item Exploration and Overlapped User Guidance (SIEOUG) for solving the MMCDR problem. SIEOUG first proposes similarity item exploration module, which not only obtains pair-wise and group-wise item-item graph knowledge, but also reduces irrelevant noise for multi-modal modeling. Then SIEOUG proposes user-item collaborative filtering module to aggregate user/item embeddings with the attention mechanism for collaborative filtering. Finally SIEOUG proposes overlapped user guidance module with optimal user matching for knowledge sharing across domains. Our empirical study on Amazon dataset with several different tasks demonstrates that SIEOUG significantly outperforms the state-of-the-art models under the MMCDR setting.

Figures (8)

• Figure 1: The problem definition of MMCDR.
• Figure 2: The model framework of proposed SIEOUG for solving MMCDR problem.
• Figure 3: The illustrations on message passing on item-item similarity graph and hypergraph. Obviously, hypergraph can aggregate more useful information during the procedure.
• Figure 4: Illustrations on overlapped user guidance. When we involve prior overlapped users as the guidance, we can avoid mismatches between users with different preferences.
• Figure 5: The experimental results on method extension.