As an important form of knowledge organization, knowledge graphs are widely recognized as one of the foundational infrastructures for the next generation of artificial intelligence technologies, receiving considerable interest from both industry and academia. Traditional methods for representing knowledge graphs mainly employ symbolic representations to explicitly describe concepts and their relationships, with clear semantics and good interpretability. However, these methods have limited coverage of knowledge types, making it challenging to apply them in open-domain scenarios. With the development of large pre-trained language models (large language models), most researchers have considered parameterized large language models as knowledge graphs. Thus, this paper focuses on the research of the life cycle of knowledge graphs in large language models. Specifically, we summarize the related work on knowledge modeling, knowledge acquisition, knowledge fusion, knowledge management, knowledge reasoning, and knowledge application. Finally, we anticipate the future development trends of large language models and knowledge graphs.
NETL
