Abstract:Knowledge graph completion (KGC) revolves around populating missing triples in a knowledge graph using available information. Text-based methods, which depend on textual descriptions of triples, often encounter difficulties when these descriptions lack sufficient information for accurate prediction-an issue inherent to the datasets and not easily resolved through modeling alone. To address this and ensure data consistency, we first use large language models (LLMs) to generate coherent descriptions, bridging the semantic gap between queries and answers. Secondly, we utilize inverse relations to create a symmetric graph, thereby providing augmented training samples for KGC. Additionally, we employ the label information inherent in knowledge graphs (KGs) to enhance the existing contrastive framework, making it fully supervised. These efforts have led to significant performance improvements on the WN18RR and FB15k-237 datasets. According to standard evaluation metrics, our approach achieves a 4.2% improvement in Hit@1 on WN18RR and a 3.4% improvement in Hit@3 on FB15k-237, demonstrating superior performance.

What problem does this paper attempt to address?

This paper attempts to solve two main problems in Knowledge Graph Completion (KGC): 1. **Mismatched Descriptions**: In text - based methods, the descriptions of entities and relations are usually collected independently, which may lead to semantic inconsistency between the query and the answer, thus affecting the prediction accuracy. For example, in the FB15k - 237 dataset, the description of the head entity "Seattle Seahawks" may lack information about "team color", causing the model to have difficulty in predicting "blue" or "gold". 2. **Pseudo - Inverses**: Current methods are too simplistic in generating inverse relations, usually just simply adding the prefix "reverse", which may confuse the model and make it difficult to accurately identify the inverse relation \(r^{-1}\), thus leading to over - fitting problems. To address these problems, the authors propose **KERMIT** (Knowledge Graph Completion of Enhanced Relation Modeling with Inverse Transformation). The specific methods are as follows: - **Generate Predictive Descriptions Using Large - Language Models (LLMs)**: By designing elaborate prompt templates (as shown in Table 1), use LLMs to generate predictive descriptions related to the query to improve the accuracy and consistency of the descriptions. - **Generate Inverse Relations**: Through another prompt template (as shown in Table 2), use LLMs to generate inverse relations to ensure the accuracy and semantic coherence of the inverse relations. - **Supervised Contrastive Learning**: Utilize the label information in the knowledge graph to extend the existing contrastive learning framework and make it a fully - supervised learning method, thereby obtaining more expressive representations. The experimental results of these methods on the WN18RR and FB15k - 237 datasets show that KERMIT significantly improves the performance of the KGC task, especially with an increase of 4.2% and 3.4% in the Hit@1 and Hit@3 metrics respectively.

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

Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion

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

KICGPT: Large Language Model with Knowledge in Context for Knowledge Graph Completion

Multi-perspective Improvement of Knowledge Graph Completion with Large Language Models

Step out of KG: Knowledge Graph Completion via Knowledgeable Retrieval and Reading Comprehension

Relational Message Passing for Fully Inductive Knowledge Graph Completion

SimKGC: Simple Contrastive Knowledge Graph Completion with Pre-trained Language Models

Relation Adversarial Network for Low Resource Knowledge Graph Completion

Knowledge Graph Completion with Pre-trained Multimodal Transformer and Twins Negative Sampling

Semantic-Enhanced Knowledge Graph Completion

Relational Message Passing for Fully Inductive Knowledge Graph Completion

Progressive Knowledge Graph Completion

Multi-level Shared Knowledge Guided Learning for Knowledge Graph Completion

From Discrimination to Generation: Knowledge Graph Completion with Generative Transformer

Knowledge Graph Completion with Information Adaptation and Refinement

Graph Structure Enhanced Pre-Training Language Model for Knowledge Graph Completion

Do Pre-trained Models Benefit Knowledge Graph Completion? A Reliable Evaluation and a Reasonable Approach.

Text-Enhanced and Relational Context Based Hyperbolic Knowledge Graph Embedding

A knowledge graph completion model based on contrastive learning and relation enhancement method