深圳大学学报（理工版）2025，Vol.42Issue(5)：597-605,9.DOI:10.3724/SP.J.1249.2025.05597
融合大语言模型与向量知识库的应用文生成框架
A framework integrating large language models and vector knowledge bases for practical document generation
摘要
Abstract
To improve the efficiency of practical document writing,this study proposes an automated generation method that integrates large language models(LLMs)and vector knowledge bases.In diverse application scenarios,manually prepared standard documents are used as templates to construct structured auxiliary files to support document generation and vector knowledge bases tailored to specific document types.By leveraging target-type document chapter headings and user-provided key information,the system queries the knowledge base to retrieve relevant text segments.Prompt engineering is then applied to guide the LLM,which synthesizes coherent text by retrieved reference segments with user input,generating content chapter by chapter in alignment the lowest-level heading structure.The generated texts are subsequently formatted into complete documents in compliance with predefined standards.Evaluation results demonstrate that,using emergency plans as a benchmark,when assessed under identical criteria with ChatGPT-4Turbo,the automatically generated emergency plans achieve 95.87%similarity to manually prepared counterparts,demonstrating comparable quality.The proposed method can overcome token limitations even under constrained computational resources,generating lengthy practical documents that meet baseline requirements.These outputs serve as reliable references or be directly adopted,significantly enhancing efficiency of document preparation.关键词
人工智能/应用文生成/大语言模型/向量知识库/提示词工程/模型评测/ChatGPT-4Turbo/DeepSeek-R1Key words
artificial intelligence/practical document generation/large language model/vector knowledge base/prompt engineering/model evaluation/ChatGPT-4Turbo/DeepSeek-R1分类
信息技术与安全科学引用本文复制引用
秦斌,陆平,徐琰,邓芳伟,王旖洋,曾渭钰,李欣莹,李灿亮..融合大语言模型与向量知识库的应用文生成框架[J].深圳大学学报（理工版）,2025,42(5):597-605,9.基金项目
University-industry Cooperation Foundation of ZTE Government Industry Large Language Model Technology Research(IA20231030016) （IA20231030016）
National Key Research and Development Program of China(2020YFB1806405) （2020YFB1806405）
Shenzhen Science and Technology Major Project(KJZD20230923114906013) ZTE政务行业大模型技术研究产学研基金资助项目(IA20231030016) （KJZD20230923114906013）
国家重点研发计划资助项目(2020YFB1806405) （2020YFB1806405）
深圳市科技重大专项资助项目(KJZD20230923114906013) （KJZD20230923114906013）
