The construction and refined extraction techniques of knowledge graph based on large language models

With the growing need for intelligent decision-support systems, the development of high-quality knowledge graphs has become essential for improving operational efficiency and decision reliability. However, the specialized nature, distributed sources, and sensitive aspects of this knowledge present unique challenges to conventional knowledge management approaches. Current general-purpose large language models often struggle with domain-specific text comprehension, particularly in accurately interpreting technical parameters and operational guidelines. To address these limitations, this paper introduces a framework for building and refining specialized knowledge graphs using adapted large language models. Our approach involves fine-tuning base LLMs with domain-specific datasets, enabling them to better handle complex terminology and semantic nuances. The framework incorporates a multimodal knowledge integration pipeline that combines rule-based systems with ontological structures to extract and link entities from diverse data sources, creating an adaptive knowledge network. Experimental results demonstrate that our fine-tuned model achieves substantial gains in relationship extraction accuracy, while the resulting knowledge graph shows strong performance in semantic coherence and operational reasoning assessments, offering robust support for critical decision-making processes. This research presents a novel approach for effective knowledge integration and cross-functional collaboration in specialized domains.