Explicit and Implicit Feature Contrastive Learning Model for Knowledge Graph Link Prediction.

Area of Science:

• Artificial Intelligence
• Data Science
• Machine Learning

Background:

• Knowledge graph link prediction is vital for inferring relationships between entities.
• Graph neural networks and contrastive learning show promise but overlook implicit entity associations.
• Existing methods struggle with distant but semantically rich entities and constrained contrastive learning.

Purpose of the Study:

• To address limitations in current knowledge graph link prediction models.
• To capture implicit associations between entities beyond direct links.
• To improve entity representations by incorporating both implicit and explicit features.

Main Methods:

• Developed an implicit feature extraction module using latent vector space clustering.
• Integrated a subgraph mechanism to preserve explicit structural information.
• Combined implicit semantic and explicit structural features for self-supervised signals.

Main Results:

• The proposed model effectively enriches entity representations by mining conceptual-level semantic features.
• The subgraph mechanism preserves crucial structural information of explicitly connected entities.
• Experimental results on benchmark datasets demonstrate superior performance over state-of-the-art baselines.

Conclusions:

• The novel approach successfully integrates implicit and explicit features for enhanced knowledge graph link prediction.
• The method captures distant, semantically rich entities previously ignored by other models.
• This work advances the field of knowledge graph construction and reasoning.