Millitary Knowledge Graph Construction Based on Universal Information Extraction Models

Data in military domain is characterized by diverse types, scattered storage and high confidentiality, which greatly limits the promotion and application of knowledge graph in military domain. The advantage of big language model in understanding complex text and semantic relations provides a possibility for the automated construction of knowledge graph in military domain. This paper proposed a knowledge graph construction method for military domain based on Universal information extraction(UIE) models, which improved the automation level of knowledge graph construction in military domain through the steps of generating high-quality corpus by domain expert annotation, unified modeling information extraction task and model fine-tuning, and provided a reference for the wide application of knowledge graph in military domain.