大模型知识管理系统  

Large Model Knowledge Management System

作　　者：周扬 蔡霈涵 董振江 ZHOU Yang;CAI Peihan;DONG Zhenjiang(Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学,南京210023

出　　处：《中兴通讯技术》2024年第2期63-71,共9页ZTE Technology Journal

基　　金：江苏省重点研发计划项目(BE2023025)。

摘　　要：企业知识管理至关重要,而传统企业知识管理系统存在构建成本高、知识利用率低的问题。提出了基于大模型检索增强生成(RAG)技术构建企业知识管理系统的方案。首先介绍了整体方案架构、业务流程与4类知识构建技术,然后重点介绍了检索前处理、知识检索、检索后处理等全流程知识检索技术,并设计了全面的测评框架。经过实践检验,该方案具有知识构建效率高且成本低、意图理解精确、知识检索准确等特点与优势。Enterprise knowledge management is very important,but traditional enterprise knowledge management systems suffer from high construction costs and low knowledge utilization rate.A scheme to build enterprise knowledge management system based on retrievalaugmented generation(RAG)technology is proposed.Firstly,the overall scheme architecture,business processes and four types of knowl⁃edge construction technologies are introduced,and then the whole process of knowledge retrieval technology are discussed,such as re⁃trieval pre-processing,knowledge retrieval,retrieval post-processing,and subsequently a comprehensive evaluation framework is de⁃signed.The scheme has the characteristics and advantages of high efficiency and low cost of knowledge construction,accurate intention un⁃derstanding,and accurate knowledge retrieval.

关 键 词：RAG 知识管理系统 大模型 知识工程 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]