Routing User-Interest Markov Tree for Scalable Personalized Knowledge-Aware Recommendation.

Area of Science:

• Artificial Intelligence
• Data Science
• Recommender Systems

Background:

• Incorporating side information, particularly from knowledge graphs (KGs), is vital for accurate and explainable recommendations.
• Existing KG-based recommendation algorithms face scalability challenges due to high-cost, hop-by-hop path enumeration strategies.
• The expanding scale of real-world data graphs necessitates more efficient KG utilization methods.

Purpose of the Study:

• To propose an end-to-end framework, KURIT-Net, that overcomes the computational and scalability limitations of traditional KG-based recommendation methods.
• To develop a novel approach that balances the routing of knowledge across both short-distance and long-distance relations within a KG.
• To enhance recommendation explainability by generating human-readable reasoning paths.

Main Methods:

• Introduced the Knowledge-tree-routed UseR-Interest Trajectories Network (KURIT-Net) framework.
• Employed user-interest Markov trees (UIMTs) to reconfigure recommendation-based KGs.
• Utilized entity and relation trajectory embedding (RTE) to summarize reasoning paths and reflect user interests.

Main Results:

• KURIT-Net effectively balances knowledge routing between entities, accommodating both proximate and distant relations.
• The framework successfully summarizes all reasoning paths within a KG, fully capturing potential user interests.
• Extensive experiments on six public datasets demonstrated that KURIT-Net significantly outperforms state-of-the-art approaches.

Conclusions:

• KURIT-Net offers a scalable and computationally efficient solution for KG-based recommendation systems.
• The proposed method provides accurate recommendations with enhanced interpretability.
• The user-interest Markov trees (UIMTs) effectively guide knowledge routing for improved recommendation quality.