深度学习在知识图谱中的最新研究现状及其分析
在撒哈拉卖雨伞1,2
1.广州医科大学 生物医学工程学院，广州番禺，511436
2.广东医科大学 生物医学工程学院，东莞松山湖，523822
摘要：知识图谱（Knowledge Graphs, KGs）作为一种结构化知识表示形式，在人工智能领域扮演着重要角色。近年来，深度学习技术与知识图谱的结合取得了显著进展，尤其在知识图谱嵌入、补全、推理以及与大语言模型（Large Language Models, LLMs）的集成等方面。本文综述了2023年至2025年深度学习在知识图谱中的最新研究现状，涵盖了主要方法、应用场景及其分析。重点讨论了图神经网络（Graph Neural Networks, GNNs）、Transformer模型在知识图谱任务中的应用，以及新兴趋势如时序知识图谱、多模态知识图谱和神经符号推理。同时，对当前挑战如幻觉问题、泛化能力和计算效率进行深入分析，并展望未来方向，包括自监督学习和多模态融合的应用。研究表明，深度学习显著提升了知识图谱的表示能力和推理效率，但仍需解决知识不完整性和可解释性问题。通过引入表格比较关键模型性能，本文提供了一个全面的框架，以指导未来研究。
关键词：知识图谱；深度学习；图神经网络；知识嵌入；知识推理；大语言模型；多模态融合
文献标志码：A 中图分类号：R318.0 doi：
Recent Research Trends and Analysis of Deep Learning in Knowledge Graphs
ZENG,Zhenhua1,2
1 School of Biomedical Engineering, Guangzhou Medical University,Guangzhou 511436,China
2. School of Biomedical Engineering, Guangdong Medical University,Dongguan 523822,China
Abstract: Knowledge Graphs (KGs), as a structured form of knowledge representation, play a vital role in the field of artificial intelligence. In recent years, the integration of deep learning techniques with knowledge graphs has achieved significant progress, particularly in areas such as knowledge graph embedding, completion, reasoning, and integration with Large Language Models (LLMs). This paper reviews the latest research status of deep learning in knowledge graphs from 2023 to 2025, covering key methods, application scenarios, and their analysis. It focuses on the application of Graph Neural Networks (GNNs) and Transformer models in knowledge graph tasks, as well as emerging trends such as temporal knowledge graphs, multimodal knowledge graphs, and neuro-symbolic reasoning. It also conducts an in-depth analysis of current challenges, including hallucination issues, generalization capabilities, and computational efficiency, while outlining future directions such as the application of self-supervised learning and multimodal fusion. Research indicates that deep learning significantly enhances the representational capacity and reasoning efficiency of knowledge graphs, though challenges like knowledge incompleteness and interpretability remain to be addressed. By introducing tabular comparisons of key model performances, this paper provides a comprehensive framework to guide future research.
Keywords: Knowledge Graph; Deep Learning; Graph Neural Networks; Knowledge Embedding; Knowledge Reasoning; Large Language Models; Multimodal Fusion
