基于检索和知识图谱的军事法规问答系统

A military legal question answering system based on retrieval and knowledge graph

军事法规对现代军事行动意义重大。针对查阅文本效率低、易出错的问题,综合运用检索和知识图谱技术构建军事法规问答系统。通过优化TF-IDF模型为法规条款建立便于逐层检索的数据结构,提升检索效率;运用知识图谱技术为法规条款生成大量知识,赋予问题理解能力;根据军事法规特点设计问句处理和条款检索算法,以短答案、条款内容和答案来源构成完整解答。分析和测试表明,构建的底层数据层次性强、语义丰富,实现的问答原型系统能够准确回答多角度提问,融合了检索与知识图谱技术的优势。本系统可为指挥员处理涉法问题提供决策辅助,所述技术同样适用于其他领域法规。

Military regulations are of great significance to modern military operations.However,consulting the texts is inefficient and error-prone.A military legal question answering system is constructed by integrating retrieval and knowledge graph technologies.The optimized TF-IDF model is used to establish a data structure for the hierarchical retrieval of laws and regulations to improve the efficiency.Knowledge graph technologies are used to generate a large amount of knowledge for the clauses of laws and regulations,ensuring the ability of question understanding.According to the characteristics of military laws and regulations,the algorithms of question processing and clause retrieval are designed,and the complete answer is composed of a short answer,clause content and an answer source.Analysis and testing show that the underlying data constructed is highly hierarchical and semantically rich.The prototype system can accurately answer questions from multiple perspectives,and it combines the advantages of retrieval and knowledge graph technologies.The system can provide decision aid for the commander to deal with legal problems,and the technologies are also applicable to regulations in other fields.