Reproduce “Multimodal Recommender Systems: A Survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qidong Liu, Jiaxi Hu, Yutian Xiao, Xiangyu Zhao, Jingtong Gao, Wanyu Wang, Qing Li, Jiliang Tang
The paper provides a comprehensive, technically-focused survey of Multimodal Recommender Systems (MRS), categorizing existing models based on technical solutions to address challenges in multimedia content understanding. It outlines a unified three-stage procedure for MRS and identifies key research directions for the field.
Key insights
- MRS models can be systematically categorized based on how they address specific technical challenges, such as multimodal feature representation, heterogeneous feature fusion, data sparsity mitigation, and model training optimization.
- Advances in computer vision and natural language processing encoders are crucial for extracting comprehensive representations from diverse multimodal inputs within MRS.
- Techniques like Disentangled Representation Learning and Contrastive Learning are increasingly used in MRS for feature enhancement, helping to distinguish unique and common characteristics across modalities and alleviate data sparsity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.