EN中文

时空信息学报 >

2025 , Vol. 32 >Issue 04: 442 - 453

DOI: https://doi.org/10.20117/j.jsti.202504001

时空赋能

基于大语言模型的土地征收业务辅助智能问答助手构建方法及实现

  • 陈卉 ,
  • 肖飞 ,
  • 咸容禹 ,
  • 郭文华 ,
  • 李正 ,
  • 徐荣强 ,
  • 胡超
展开
  • 1.自然资源部信息中心,北京 100830;
    2.浙江时空智子大数据有限公司,宁波 315200
陈卉,研究方向为政府网站建设、智能化技术应用等。E-mail:hchen@infomail.mnr.gov.cn
咸容禹,研究方向为自然资源政务信息化系统建设和管理等。E-mail:ryxian@mail.mnr.gov.cn

收稿日期: 2025-04-01

  修回日期: 2025-07-02

  网络出版日期: 2026-05-06

基金资助

国家重点研发计划项目（2023YFC3804005）

收起

Technical architecture for constructing an intelligent Q&A assistant for land expropriation operations based on large language models

  • CHEN Hui ,
  • XIAO Fei ,
  • XIAN Rongyu ,
  • GUO Wenhua ,
  • LI Zheng ,
  • XU Rongqiang ,
  • HU Chao
Expand
  • 1. Information Center, the Ministry of Natural Resources, Beijing 100830, China;
    2. Zhejiang Spatiotemporal Sophon Bigdata Co., Ltd., Ningbo 315200, China

Received date: 2025-04-01

  Revised date: 2025-07-02

  Online published: 2026-05-06

Fold

摘要

针对土地征收业务中存在的知识碎片化、响应滞后与专业性不足等问题,以土地征收业务辅助问答智能助 手建设为例,本文构建一种基于大语言模型的土地征收业务辅助智能问答助手的技术框架,以突破传统系统在统 一知识组织与高效问答支持方面的能力瓶颈。通过构建业务知识体系与多维知识库、设计检索增强生成机制、搭 建具备意图识别与任务拆解能力的智能体架构,并结合三维评测集与提示词优化方法,实现对复杂土地征收业务 问题的高效应答;进一步研发形成土地征收业务辅助智能问答助手系统,为验证方法可行性,并进行专家评测对 比分析。结果表明：与基础模型相比,本文方法在专业性、实用性方面均实现显著提升,尤其在政策文本生成及 费用测算逻辑组织等较复杂任务中,答复内容更为系统、准确且具操作性,可为提升土地征收业务办理效率提供 助力;本文方法整体得分提升了 50.8%,复杂问题场景下的得分提升超过了 42.1%。

关键词： 大语言模型; 土地征收; 智能问答; 业务场景; 增强优化路径; 数据治理

本文引用格式

陈卉 , 肖飞 , 咸容禹 , 郭文华 , 李正 , 徐荣强 , 胡超 . 基于大语言模型的土地征收业务辅助智能问答助手构建方法及实现[J]. 时空信息学报, 2025 , 32(04) : 442 -453 . DOI: 10.20117/j.jsti.202504001

Abstract

[Objective] The rapid advancement of large language model (LLM) technology has unlocked unprecedented capabilities in natural language understanding and generation, creating new paradigms for government digital transformation and the delivery of efficient public services. Within this context, land acquisition—a domain of significant public interest among natural resource governance issues—persistently confronts challenges such as fragmented knowledge bases and delayed responses to citizen inquiries. Conventional question-answering systems often fail to meet the dual demands of high accuracy and timeliness required in practical operations. This paper addresses these challenges by proposing a specialized technical framework founded on LLM for constructing an intelligent assistant designed to support land acquisition professionals. The research objective is to develop a mission-oriented system featuring authoritative domain knowledge, robust professional reasoning abilities, and sophisticated knowledge fusion capabilities, thereby addressing the efficiency, expertise, and intelligence requirements inherent in land administration workflows.
[Method] We introduce a novel intelligent question-answering framework tailored for the complexities of land acquisition business. This framework prioritizes three core components: systematic knowledge engineering, advanced query processing, and comprehensive system evaluation. First, we establish a multidimensional knowledge repository encompassing policies, regulations, procedural workflows, and computational rules. This structure enables the unified organization and efficient retrieval of previously disparate information. Second, to enhance knowledge retrieval and improve response precision, we design a targeted segment extraction method. This approach leverages semantic matching and structural localization to ground generated answers firmly within source material. Architecturally, the system employs a modular design, manifested as an intelligent agent capable of intent recognition, task decomposition, and result synthesis. For performance validation, we implement a three-dimensional evaluation scheme that assesses the system across diverse problem types, levels of complexity, and modes of expression. Furthermore, we incorporate a multidimensional prompt optimization strategy to stabilize and refine language generation in complex task scenarios. This integrated methodology offers a scalable technical pathway for applying LLM to intricate professional fields like land acquisition within the natural resources sector.
[Result] Based on the proposed methodology, we developed a functional intelligent Q&A assistant for land acquisition support and conducted a comparative evaluation against a baseline LLM. The results demonstrate substantial improvements in professionalism, practical utility, and the capacity to handle complex queries. Particularly in demanding tasks such as policy document drafting and cost estimation, the system exhibited superior logical structuring and actionable output. Compared to the baseline model, our system achieved an overall performance increase of 50.8%. Notably, its effectiveness in complex problem scenarios showed a gain exceeding 42.1%.
[Conclusion] The intelligent Q&A assistant framework presented in this study effectively mitigates the challenges of implementing intelligent responses in land acquisition processes. The architecture demonstrates strong reusability and scalability, offering a verifiable and adaptable technical solution for enhancing intelligent Q&A functions within large language models operating in the natural resources industry. These findings provide a valuable reference for facilitating the low-cost deployment and efficient application of LLM in representative industrial contexts.

Key words： large language model; land expropriation; intelligent Q&A; business scenarios; enhance the optimization path; data governance

参考文献

陈冬雷, 秦薇.2024.基于检索增强生成的智能问答系统关键技术研究.办公自动化, 29(19):82-86[Chen D L, Qin W.2024.Research on the key technologies for intelligent Q & a system based on retrieval augmented generation technology.Office Informatization, 29(19):82-86 (in Chinese)]
陈光, 郭军.2024.大语言模型时代的人工智能:技术内涵、行业应用与挑战.北京邮电大学学报, 47(4):20-28[Chen G, Guo J.2024.Artificial intelligence in the era of large language models:Technical significance, industry applications, and challenges.Journal of Beijing University of Posts and Telecommunications, 47(4):20-28 (in Chinese)]
陈军, 王艳慧, 武昊, 刘万增.2023.时空信息赋能高质量发展的基本问题与发展方向.时空信息学报, 30(1):1-11[Chen J, Wang Y H, Wu H, Liu W Z.2023.Basic issues and development directions of high-quality development empowered by spatio-temporal information.Journal of Spatio-temporal Information, 30(1):1-11 (in Chinese)]
陈文泰, 张帆.2025.提示词构成要素对大语言模型跨模态内容生成质量的影响研究——基于讯飞星火大模型文生图功能的探索性实验.郑州大学学报 ( 哲学社会科学版 ), 58(2):99-104[Chen W T, Zhang F.2025.Research on the influence of hint elements on the quality of cross-modal content generation of large language model—An exploratory experiment based on the function of iFLYTEK spark model's literatures and graphs.Journal of Zhengzhou University (Philosophy and Social Sciences Edition), 58(2):99-104 (in Chinese)]
陈展鹏, 杜启勇, 胡鑫, 杨学习, 王天应, 江一凡, 尹姝彤, 邹煜星.2025.面向辅助用地报批的知识图谱协同构建与智能问答方法及实现.时空信息学报, 32(1):94-103[Chen Z P, Du Q Y, Hu X,Yang X X, Wang T Y, Jiang Y F, Yin S T, Zou Y X.2025.Methods and implementation of collaborative construction and intelligent question-answering for knowledge graphs in support of land use approval.Journal of Spatio-temporal Information, 32(1):94-103 (in Chinese)]
程大伟, 吴佳璇, 李江彤, 丁志军, 蒋昌俊.2025.大模型金融场景能力评测框架研究.计算机科学, 52(3):239-247[Cheng D W, Wu J X, Li J T, Ding Z J, Jiang C J.2025.Study on evaluation framework of large language model's financial scenario capability.Computer Science, 52(3):239-247 (in Chinese)]
邓建鹏, 赵治松.2025.DeepSeek 的破局与变局:论生成式人工智能的监管方向.新疆师范大学学报( 哲学社会科学版), 46(4):99-108[Deng J P, Zhao Z S.2025.The breakthrough and transformation of DeepSeek:The new rules for AIGC regulation.Journal of Xinjiang Normal University (Philosophy and Social Sciences), 46(4):99-108 (in Chinese)]
段淳林, 蒲源, 魏方.2025.智能体与生成式计算广告创意的模型建构与效果研究.新闻界,1-12.https://link.cnki.net/doi/10.15897/j.cnki.cn51-1046/g2.20250603.001.[2025-06-04][Duan C L, Pu Y,Wei F.2025.The model construction and effects analysis of agent and generative computational advertising creatives.Journalism and Mass Communication,1-12.https://link.cnki.net/doi/10.15897/j.cnki.cn51-1046/g2.20250603.001.[2025-06-04] (in Chinese)]
房绍坤.2019.土地征收制度的立法完善——以《土地管理法修正案草案》为分析对象.法学杂志, 40(4):1-12[Fang S K.2019.Legal perfection of land expropriation system:Taking the draft amendment of land administration law as the object of analysis.Law Science Magazine, 40(4):1-12 (in Chinese)]
高雯, 张轩铭, 杨家文, 林波荣, 刘洋, 陈宛晴, 杜修力.2025.基于大语言模型的建筑法规智能问答平台研究.建筑技艺(中英文), 31(2):53-57[Gao W, Zhang X M, Yang J W, Lin B R, Liu Y, Chen W Q, Du X L.2025.Research on intelligent Q & A platform for building regulations based on large language model.Architecture Technique, 31(2):53-57 (in Chinese)]
龚为纲, 黄思源.2025.大语言模型与计算社会科学:工具、议题与挑战.求索,(3):137-148[Gong W G, Huang S Y.2025.Large language models and computational social science:Tools, issues and challenges.Seeker, (3):137-148 (in Chinese)]
侯一俊, 杨征, 陈卉, 谢敏, 徐晓婧, 马连刚.2024.政府门户网站政民互动研究——基于自然资源部门户网站公众留言的实例分析.自然资源信息化,(5):47-53, 87[Hou Y J, Yang Z, Chen H,Xie M, Xu X J, Ma L G.2024.Research on government-citizen interaction on government portal websites:A case analysis based on public comments on the portal website of Ministry of Natural Resources.Natural Resources Informatization,(5):47-53, 87 (in Chinese)]
胡伟, 姜晓夏, 邵洲天, 朱容志.2024.大型语言模型智能体:机制和应用的综述.指挥信息系统与技术, 15(6):1-11[Hu W, Jiang X X, Shao Z T, Zhu R Z.2024.Large language model(LLM) agents:Survey of mechanisms and applications.Command Information System and Technology, 15(6):1-11 (in Chinese)]
刘伟, 谢海斌, 陈少飞.2024.Agent 视域下的人工智能赋能作战系统.指挥控制与仿真, 46(6):8-14[Liu W, Xie H B, Chen S F.2024.Artificial intelligence empowered combat systems from the perspective of Agent.Command Control & Simulation, 46(6):8-14 (in Chinese)]
刘雪颖, 云静, 李博, 史晓国, 张钰莹.2025.基于大型语言模型的检索增强生成综述.计算机工程与应用, 61(13):1-25[Liu X Y, Yun J, Li B, Shi X G, Zhang Y Y.2025.Survey of retrieval-augmented generation based on large language models.Computer Engineering and Applications, 61(13):1-25 (in Chinese)]
彭文祥, 张德英.2024.基于 LSTM 与 Transformer 的地面沉降智能预测方法研究 —— 以上海市为例.时空信息学报, 31(1):94-103[Peng W X, Zhang D Y.2024.Research on land subsidence intelligent prediction method based on LSTM and Transformer:A case study of Shanghai.Journal of Spatio-temporal Information, 31(1):94-103 (in Chinese)]
饶东宁, 许正辉, 梁瑞仕.2025.基于知识库问答的回答生成研究.计算机工程, 51(2):94-101[Rao D N, Xu Z H, Liang R S.2025.Research on answer generation based on knowledge base question answering.Computer Engineering, 51(2):94-101 (in Chinese)]
王东清, 芦飞, 张炳会, 李道童, 彭继阳, 王兵, 姚藩益, 艾山彬.2025.大语言模型中提示词工程综述.计算机系统应用, 34(1):1-10[Wang D Q, Lu F, Zhang B H, Li D T, Peng J Y, Wang B, Yao F Y, Ai S B.2025.Survey on prompt engineering in large language model.Computer Systems and Applications, 34(1):1-10 (in Chinese)]
王海军.2024.矿山行业大模型建设路径探索与应用展望.煤炭科学技术, 52(11):45-59[Wang H J.2024.Construction exploration and application prospect of the large model in mining industry.Coal Science and Technology, 52(11):45-59 (in Chinese)]
王益国, 杨洁, 李馥孜, 郭萍萍.2024.基于 AI Agent 的高校图书推荐架构研究.大学图书情报学刊, 42(6):3-8[Wang Y G, Yang J, Li F Z, Guo P P.2024.Research on university library book recommendation architecture based on AI agent.Journal of Academic Library and Information Science, 42(6):3-8 (in Chinese)]
吴璇, 付涛.2025.检索增强生成技术研究综述.计算机工程与应用, 1-20.https://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSGG20250610007&dbname=CJFD&dbcode=CJFQ.[2025-06-11][Wu X, Fu T.2025.Comprehensive review of retrieval-augmented generation.Computer Engineering and Applications,1-20.https://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSGG20250610007&dbname=CJFD&dbcode=CJFQ.[2025-06-11] (in Chinese)]
肖飞, 吴洪涛, 郭文华, 刘文超, 曾建鹰, 熊伟, 张芸, 胡源, 沈坚,董洁飞, 陈志远.2024.自然资源行业大模型建设框架与典型场景研究.自然资源信息化,(6):54-62[Xiao F, Wu H T, Guo W H, Liu W C, Zeng J Y, Xiong W, Zhang Y, Hu Y, Shen J, Dong J F, Chen Z Y.2024.Research on the framework and typical scenarios of large language model construction in the natural resources industry.Natural Resources Informatization, (6):54-62 (in Chinese)]
谢敏, 叶兴茂, 侯一俊, 徐晓婧, 赖敏.2020.我国征地信息公开信息平台建设回顾与展望.国土资源信息化,(2):20-25[Xie M, Ye X M, Hou Y J, Xu X J, Lai M.2020.Review and prospect of the construction of opening land requisition information platform in China.Land and Resources Informatization, (2):20-25 (in Chinese)]
徐晓东, 全喜伟, 杨帅.2025.大语言模型在交管政务服务中的应用研究.道路交通管理, (3):36-39[Xu X D, Quan X W, Yang S.2025.Research on the application of large language model in traffic control government service.Road Traffic Management, (3):36-39 (in Chinese)]
徐子阳, 周绍光, 葛莹, 万子豪.2025.基于模型蒸馏的无标签城市道路遥感自动提取.时空信息学报, 32(2):113-126[Xu Z Y, Zhou S G, Ge Y, Wan Z H.2025.Automatic extraction of unlabeled urban roads from remote sensing images based on model distillation.Journal of Spatio-temporal Information, 32(2):113-126 (in Chinese)]
张浩然, 郝文宁, 靳大尉, 程恺, 翟颖.2025.DF-RAG:基于查询重写和知识选择的检索增强生成方法.计算机科学, 1-12.[https://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSJA2025022600E&dbname=CJFD&dbcode=CJFQ.[2025-02-27][Zhang H R, Hao W N, Jin D W, Cheng K, Zhai Y.2025.DF-RAG:Search enhancement generation method based on query rewriting and knowledge selection.Computer Science,1-12.https://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSJA2025022600E&dbname=CJFD&dbcode=CJFQ.[2025-02-27](in Chinese)]
张金华, 仇培元, 邢华桥, 冯永玉.2024.面向地表覆盖处理服务的知识图谱建模方法研究.时空信息学报, 31(5):641-651[Zhang J H, Qiu P Y, Xing H Q, Feng Y Y.2024.A method for construction of knowledge graph for land cover processing services.Journal of Spatio-temporal Information, 31(5):641-651 (in Chinese)]
张金营, 王天堃, 么长英, 谢华, 柴林政, 刘书恺, 李彤亮, 李舟军.2024.基于大语言模型的电力知识库智能问答系统构建与评价.计算机科学, 51(12):286-292[Zhang J Y, Wang T K, Yao C Y, Xie H, Chai L Z, Liu S K, Li T L, Li Z J.2024.Construction and evaluation of intelligent question answering system for electric power knowledge base based on large language model.Computer Science, 51(12):286-292 (in Chinese)]
张晏,吴孔逸.2024.以自然资源数字化发展助力实现中国式现代化.中国自然资源报,2024-02-07(002)[Zhang Y, Wu K Y.2024.To promote the realization of Chinese path to modernization through the digital development of natural resources.China Natural Resources News, 2024-02-07 (002) (in Chinese)]
赵浜, 曹树金.2024.生成式 AI 大模型结合知识库与 AI Agent 开展知识挖掘的探析.图书情报知识,1-14.https://kns.cnki.net/KCMS/detail/detail.aspx?filename=TSQC20241102001&dbname=CJFD&dbcode=CJFQ.[2024-11-04][Zhao B, Cao S J.2024.Exploring knowledge mining using large language model combined with knowledge base and AI agent.Documentation, Information&Knowledge,1-14.https://kns.cnki.net/KCMS/detail/detail.aspx?filename=TSQC20241102001&dbname=CJFD&dbcode=CJFQ.[2024-11-04] (in Chinese)]
Options
文章导航

/

〈 〉