Military question and answer method and system for large language model-driven knowledge graph multi-hop reasoning

The invention discloses a large language model driven knowledge graph multi-hop reasoning military question and answer method and system, and the method comprises the steps: giving a military knowledge graph and a large language model, and inputting a military domain question; reasoning an initial entity set through a large language model; reasoning chain exploration circulation, including center adjacent domain relation exploration, center adjacent domain relation clipping, center adjacent domain entity exploration and center adjacent domain entity clipping; inference path evaluation: evaluating whether the inference path meets the semantic requirement of the question, if not, continuing an inference chain exploration cycle, and otherwise, determining a chain tail entity as an answer; understanding is assisted through a large language model, and logic analysis of answers and the reasoning process is given. The system comprises a starting entity reasoning module, a reasoning chain exploration circulation modu