Complete the exploration of low-resource knowledge graph completion based on large model technology

In the construction and application of knowledge graph, it is a realistic research problem to complete the knowledge in the low resource field. Traditional methods that rely on manual annotation and rules are not only costly, but also have limitations in coverage and scalability. To solve this problem, this paper proposes a large model technique, combining fine-tuning and knowledge transfer strategies. Firstly, to improve the ability of fine-tuning of the large model, the rich knowledge learned by the large model in the high resource field to assist the completion of the low resource knowledge graph through knowledge transfer technology to make up for the shortage of direct extraction. The experimental results show that this method can effectively improve the completion rate and accuracy of the knowledge graph, especially in the completion of entity relations and attribute filling. Furthermore, we explore the impact of different fine-tuning strategies and knowledge transfer methods on the completion effect, providing experimental empirical and theoretical support for future studies on similar issues.