Abstract:Knowledge graphs (KGs) serve as structured representations of knowledge, comprising entities and relations. KGs are inherently incomplete, sparse, and have a strong need for completion. Although many knowledge graph embedding models have been designed for knowledge graph completion, they predominantly focus on capturing observable correlations between entities. Due to the sparsity of KGs, potential semantic correlations are challenging to capture. To tackle this problem, we propose a model entitled semantic-enhanced knowledge graph completion (SE-KGC). SE-KGC effectively addresses the issue by incorporating predefined semantic patterns, enabling the capture of semantic correlations between entities and enhancing features for representation learning. To implement this approach, we employ a multi-relational graph convolution network encoder, which effectively encodes the KG. Subsequently, we utilize a scoring decoder to evaluate triplets. Experimental results demonstrate that our SE-KGC model outperforms other state-of-the-art methods in link-prediction tasks across three datasets. Specifically, compared to the baselines, SE-KGC achieved improvements of 11.7%, 1.05%, and 2.30% in terms of MRR on these three datasets. Furthermore, we present a comprehensive analysis of the contributions of different semantic patterns, and find that entities with higher connectivity play a pivotal role in effectively capturing and characterizing semantic information.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper aims to address the issues of incompleteness and sparsity in Knowledge Graphs (KGs). Specifically, although many Knowledge Graph Embedding (KGE) models have been designed for knowledge graph completion, these models mainly focus on capturing observable associations between entities while ignoring potential semantic associations. Due to the sparsity of knowledge graphs, potential semantic associations are difficult to capture, which limits the performance of existing models in the task of knowledge graph completion. To solve this problem, the authors propose a new model—Semantic-Enhanced Knowledge Graph Completion (SE-KGC). This model introduces predefined semantic patterns to effectively capture semantic associations between entities and enhance feature representation learning. Specifically, SE-KGC uses a multi-relational graph convolution network to encode the knowledge graph and employs a scoring decoder to evaluate triples. Experimental results show that SE-KGC outperforms existing state-of-the-art methods in link prediction tasks on multiple datasets, particularly achieving significant improvements in the MRR metric. Additionally, the authors conduct a comprehensive analysis of the contributions of different semantic patterns, finding that entities with higher connectivity play a key role in effectively capturing and representing semantic information. ### Main Contributions 1. **Proposing the SE-KGC Model**: This model can capture potential semantic associations between entities. 2. **Developing an Entity Feature Enhancement Module**: This module combines semantic and structural information to adaptively enrich the local neighborhood of sparse knowledge graphs for use by the graph convolution network encoder. 3. **Proving Effectiveness through Experiments**: Experiments on multiple real-world datasets demonstrate the effectiveness of SE-KGC, with an in-depth analysis of the learning weights of different semantic patterns, revealing that entities with higher connectivity are more important. ### Key Technologies of the Solution 1. **Semantic Enhancement**: Capturing potential semantic associations through predefined semantic patterns (such as triangles, rectangles, etc.). 2. **Multi-Relational Graph Convolution Network**: Used to learn robust local neighborhood representations. 3. **Scoring Decoder**: Used to evaluate the validity of triples. ### Experimental Results Experimental results show that SE-KGC outperforms existing state-of-the-art methods in link prediction tasks on three datasets, particularly improving the MRR metric by 11.7%, 1.05%, and 2.30%, respectively. Additionally, the analysis of the contributions of different semantic patterns indicates that entities with higher connectivity play a key role in capturing and representing semantic information. ### Conclusion By introducing a semantic enhancement module, SE-KGC effectively addresses the issues of incompleteness and sparsity in knowledge graphs, improving the performance of knowledge graph completion tasks. The model's performance on multiple datasets demonstrates its effectiveness and competitiveness.

Semantic-Enhanced Knowledge Graph Completion

Learning Context-based Embeddings for Knowledge Graph Completion

Semantic- and relation-based graph neural network for knowledge graph completion

Leveraging Lexical Semantic Information for Learning Concept-Based Multiple Embedding Representations for Knowledge Graph Completion.

Semantic Interaction Matching Network for Few-shot Knowledge Graph Completion

A semantic guide‐based embedding method for knowledge graph completion

SEPAKE: a structure-enhanced and position-aware knowledge embedding framework for knowledge graph completion

GS-InGAT: An interaction graph attention network with global semantic for knowledge graph completion

Towards Semantically Enriched Embeddings for Knowledge Graph Completion

A Comprehensive Study on Knowledge Graph Embedding over Relational Patterns Based on Rule Learning

Federated Knowledge Graph Completion Via Embedding-Contrastive Learning

Knowledge Graph Completion Method of Combining Structural Information with Semantic Information

GLSEC: Global and Local Semantic-Enhanced Contrastive Framework for Knowledge Graph Completion

Retrieval, Reasoning, Re-ranking: A Context-Enriched Framework for Knowledge Graph Completion

Relation-Aware Multi-Positive Contrastive Knowledge Graph Completion with Embedding Dimension Scaling

Hierarchical Perceptual Graph Attention Network for Knowledge Graph Completion

KERMIT: Knowledge Graph Completion of Enhanced Relation Modeling with Inverse Transformation

Multi-level Shared Knowledge Guided Learning for Knowledge Graph Completion

GLSEC:Global and local semantic-enhanced contrastive framework for knowledge graph completion

Knowledge graph completion method based on quantum embedding and quaternion interaction enhancement

Progressive Knowledge Graph Completion