基于大模型提取用户意图结合变分自编码器与对比学习的序列推荐方法
By combining pruning and masking sequence enhancement techniques, variational autoencoders, and large language models, the limitations of semantic loss and long-term dependency modeling in sequence recommendation are overcome, improving the robustness and accuracy of the recommendation system. The generated recommendation list is diverse and interpretable.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NORTHEASTERN UNIV CHINA
- Filing Date
- 2025-11-27
- Publication Date
- 2026-07-17
AI Technical Summary
Existing sequence recommendation methods based on contrastive learning have limitations in long-term dependency modeling, face the problem of state space explosion, and excessive data augmentation leads to semantic loss, reducing the accuracy and robustness of the model.
Different data views are generated by using pruning and masking sequence enhancement techniques, combined with variational autoencoder (VAE) for semantic reconstruction and repair, and contrastive learning is performed in the latent space to construct multi-task learning objectives. A large language model is introduced to understand user intent, and recommendation sequences are generated by inverse sorting fusion.
It effectively alleviates the semantic loss problem, improves the robustness and accuracy of the model in sparse data environments, enhances the generalization ability and interpretability of the recommendation system, and generates recommendation lists that are diverse, accurate and relevant.
Smart Images
Figure CN121579781B_ABST