Multi-intent Aware Contrastive Learning for Sequential Recommendation

Junshu Huang,Zi Long,Xianghua Fu,Yin Chen
2024-09-13
Abstract:Intent is a significant latent factor influencing user-item interaction sequences. Prevalent sequence recommendation models that utilize contrastive learning predominantly rely on single-intent representations to direct the training process. However, this paradigm oversimplifies real-world recommendation scenarios, attempting to encapsulate the diversity of intents within the single-intent level representation. SR models considering multi-intent information in their framework are more likely to reflect real-life recommendation scenarios accurately.
Machine Learning
What problem does this paper attempt to address?