Cross-modal contrastive learning for aspect-based recommendation

Heesoo Won, Byungkook Oh, Hyeongjun Yang, Kyong-Ho Lee

Abstract: Highlights•Integrate multimodal data (knowledge graph and review text) for aspect-based recommendation.•Propose a novel cross-modal contrastive learning with the correlation between multimodal data.•Learn informative de-biased aspect-level representation by comparing signals from multimodal data.