Personalized scenic spot recommendation method based on multi-modal knowledge graph comparative learning

By constructing a multimodal knowledge graph and utilizing embedding models of visual, textual, and structured information, combined with knowledge-aware graph convolutional neural networks and graph attention networks, the problems of insufficient multimodal information fusion, inadequate personalized modeling, and insufficient robustness in existing tourism recommendation systems are solved, thus realizing personalized and timely attraction recommendations.

CN121658719AActive Publication Date: 2026-03-13OCEAN UNIV OF CHINA
3 Cites 3 Cited by

Patent Information

Application Number
CN202511852028.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-10
Publication Date
2026-03-13
Estimated Expiration
2045-12-10

AI Technical Summary

Technical Problem

Existing tourism recommendation systems have shortcomings in multimodal information fusion, personalized structural modeling, robustness in cold start scenarios, and dynamic capture of user interests, resulting in insufficient interpretability and timeliness of recommendation results.

Method used

By constructing a multimodal knowledge graph, and combining visual feature embedding, text feature embedding, and knowledge-aware graph convolutional neural network with structure-aware contrastive learning and graph attention network, we can achieve adaptive fusion of multimodal information and dynamic user interest modeling, thereby improving the robustness and personalization of the recommendation system.

Benefits of technology

It achieves comprehensive modeling of scenic spot features, enhances the comprehensiveness and personalization of recommendation results, improves the robustness of the model and the timeliness of recommendations, and can dynamically capture user interests to generate more accurate personalized recommendations.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of artificial intelligence and intelligent recommendation, and particularly discloses a personalized scenic spot recommendation method based on multi-modal knowledge graph comparative learning, which comprises the following steps: constructing a tourism field-oriented multi-modal knowledge graph, and then respectively establishing a visual feature embedding module and a text feature embedding module; then, modeling is carried out on a graph structure containing user interests through a knowledge perception graph convolutional network, user preferences, entity semantics and relation semantics information are fused, and structured knowledge embedding is obtained; introducing a structure-perceived comparative learning mechanism, and carrying out importance evaluation and enhanced disturbance on the graph structure; generating a unified multi-modal scenic spot representation through a gating fusion network; and outputting a personalized recommendation result. The multi-modal information is effectively fused under the conditions of weak supervision and data sparseness, high-precision scenic spot recommendation is realized, the personalization and robustness of a recommendation system are improved, and the method is suitable for application scenes such as intelligent travel, travel navigation and cultural scenic spot recommendation.
Need to check novelty before this filing date? Find Prior Art