Graph attention and text semantics improve personalized recommendation.

Area of Science:

• Computer Science
• Artificial Intelligence
• Information Retrieval

Background:

• Information overload necessitates effective personalized recommendation systems.
• Traditional collaborative filtering methods face challenges with data sparsity and cold-start problems.
• Existing models struggle to capture complex user interests and item relationships.

Purpose of the Study:

• To propose a novel personalized recommendation model integrating graph attention mechanisms and auxiliary textual information.
• To address data sparsity and cold-start issues in recommendation systems.
• To enhance recommendation accuracy and robustness through semantic enrichment.

Main Methods:

• Modeling user-item interactions on a knowledge graph.
• Employing graph attention networks (GANs) for multi-hop user interest capture.
• Utilizing graph convolution for item-side neighborhood aggregation.
• Encoding textual data into semantic embeddings to enrich graph entity representations.

Main Results:

• The proposed model demonstrated superior performance on AUC, F1-score, and Top-K recall.
• Achieved better results compared to state-of-the-art baseline methods.
• Effectively improved recommendation accuracy and robustness on sparse datasets.

Conclusions:

• Combining graph-based modeling with textual semantic enhancement significantly boosts recommendation performance.
• The model offers a robust solution for personalized recommendations in data-scarce environments.
• This approach effectively captures intricate user preferences and item characteristics.