Recently, the research team led by Professor Qian Shenyi from the School of Computer Science and Technology published a review entitled “A review of background, methods, limitations and opportunities of knowledge graph completion” in Computer Science Review (Chinese Academy of Sciences Q1 TOP journal, IF = 12.7), a high-quality journal under Elsevier.

With the rapid development of artificial intelligence technology, knowledge graphs, as the core technology for structured knowledge representations, play a significant role in multiple fields such as natural language processing, recommendation systems, and intelligent question answering. However, the widespread existence of incomplete information in real-world applications of knowledge graphs has weakened the practical utility. Knowledge graph completion technology, which aims to automatically infer and supplement the missing entities or relations, has become a critical means to improve the quality and usability of knowledge graphs.

Traditional closed-domain knowledge graph completion mainly relies on embedding or path models, which performs well on structured data yet demonstrates limited capability in handling unseen entities and relations in open scenarios. In recent years, breakthroughs in neural networks and large language models have spurred the emergence of open-domain knowledge graph completion. However, there is still a lack of systematic analysis and classification of model architectures. Therefore, this paper systematically reviews the basic theories and methodological systems in the field of knowledge graphs, and proposes a unified model classification framework spanning both closed and open domains, which classifies existing methods into four categories, namely the embedding-based, path-based, neural network-based, and large language model-based methods.

By integrating the mainstream dataset resources in the field, the paper engages in a comprehensive comparison and analysis of various methods from multiple perspectives. It further explores the current challenges and future development directions of the knowledge graph completion technology, including key issues such as the complex knowledge reasoning, the improvement of domain adaptation ability, and the deep integration of large language models with knowledge graphs. These efforts provide a theoretical basis and practical reference for subsequent research.

This paper was supported by multiple scientific research projects, including the National Natural Science Foundation of China, the Major Project for Public Welfare of Henan Province, and the Henan Province Science and Technology Program for Tackling Key Problems.

Journal article link: https://doi.org/10.1016/j.cosrev.2025.100809