基于多智能体层次化协同的地理事件抽取与时空解析

doi:10.11947/j.AGCS.2025.20250183

测绘学报 ›› 2025, Vol. 54 ›› Issue (11): 2052-2067.doi: 10.11947/j.AGCS.2025.20250183

• 地图学与地理信息 • 上一篇

基于多智能体层次化协同的地理事件抽取与时空解析

胡鑫¹(), 杨学习¹^,²(), 江一凡¹, 王宪彬¹, 丁晨³, 谢顾然¹, 邓敏¹^,²^,⁴

^1.中南大学地球科学与信息物理学院，湖南　长沙　410083
^2.湖南省地理空间信息工程技术研究中心，湖南　长沙　410018
^3.华南师范大学北斗研究院，广东　佛山　528225
^4.江西师范大学地理与环境学院，江西　南昌　330022

收稿日期:2025-04-28 修回日期:2025-10-16 发布日期:2025-12-15
通讯作者: 杨学习 E-mail:225001019@csu.edu.cn;yangxuexi@csu.edu.cn
作者简介:胡鑫（1999—），男，博士生，研究方向为时空知识图谱构建与应用。E-mail：225001019@csu.edu.cn
基金资助:
国家重点研发计划(2022YFB3904203);国家自然科学基金(42271485);湖南省自然科学基金(2024JJ1009);中南大学前沿交叉项目(2023QYJC002);江西省双千计划(jxsq2020102062);湖南省地理空间信息工程技术研究中心开放课题(HNGIET2024006)

Hierarchical multi-agent collaboration for geographic event extraction and spatio-temporal parsing

Xin HU¹(), Xuexi YANG¹^,²(), Yifan JIANG¹, Xianbin WANG¹, Chen DING³, Guran XIE¹, Min DENG¹^,²^,⁴

^1.School of Geosciences and Info-Physics, Central South University, Changsha 410083, China
^2.Hunan Geospatial Information Engineering Technology Research Center, Changsha 410018, China
^3.Beidou Research Institute, South China Normal University, Foshan 528225, China
^4.School of Geography and Environment, Jiangxi Normal University, Nanchang 330022, China

Received:2025-04-28 Revised:2025-10-16 Published:2025-12-15
Contact: Xuexi YANG E-mail:225001019@csu.edu.cn;yangxuexi@csu.edu.cn
About author:HU Xin (1999—), male, PhD candidate, majors in construction and application of spatio-temporal knowledge graph.　E-mail: 225001019@csu.edu.cn
Supported by:
The National Key Research and Development Program of China(2022YFB3904203);The National Natural Science Foundation of China(42271485);The Natural Science Foundation of Hunan Province, China(2024JJ1009);The Frontier Interdisciplinary Project of Central South University(2023QYJC002);The Double-Thousand Plan of Jiangxi Province(jxsq2020102062);The Open Project of Hunan Geospatial Information Engineering Technology Research Center(HNGIET2024006)

摘要/Abstract

摘要：

从海量文本中精准捕获与解析地理事件，对于深刻理解现实世界动态变化、构建富含时空语义的时空知识图谱至关重要。然而，描述时空位置的形式多样且内涵模糊，同时领域内高质量标注样本匮乏，这给地理事件信息的准确提取带来了严峻挑战。对此，本文提出一种多智能体层次化协同方法，系统地将大语言模型的推理能力应用于少样本场景下的地理事件抽取与解析任务。本文方法的核心在于构建了一个由主控协调智能体与多个专精子智能体组成的协作框架，主控智能体负责对任务进行自适应分解与调度，专精智能体则专注于地理事件抽取、时间要素解析、空间要素定位及空间位置推理等子任务。在基于DUEE构建的基础数据集和空间推理增强数据集上的试验表明，本文方法在少样本条件下展现出优异的地理事件时空解析性能。其中，本文方法在基础数据集上的空间要素解析（F₁@100 m=0.779）与时间要素解析（F_1time=0.856）性能，相较于当前最优基线方法分别实现了16.3%与22.1%的显著提升。同时，消融研究也验证了该方法设计的有效性。此外，通过对“7·20郑州特大暴雨”事件相关社交媒体数据的实例分析，进一步证实了本文方法在解析事件时空发展脉络、辅助理解事件演化过程方面的能力。本文研究为面向文本数据的地理事件抽取提供了一种多智能体处理范式，其成果有望为事件驱动的时空知识图谱及地理空间智能提供有力支撑。

关键词: 地理事件, 事件抽取, 时空要素解析, 多智能体, 大语言模型

Abstract:

Accurately extracting and interpreting geographic events from vast amounts of text is critical for understanding real-world dynamics and constructing semantically rich spatio-temporal knowledge graph (STKG). However, the diverse expressions and inherent ambiguities of spatio-temporal information within these texts present significant challenges. To address these challenges, this paper proposes a multi-agent hierarchical collaborative method that systematically applies the reasoning capabilities of large language models (LLMs) to the task of geographic event extraction and parsing in few-shot scenarios. The core of our approach is a collaborative framework consisting of a master agent and multiple specialized sub-agents. The master agent performs adaptive task decomposition and scheduling, while the specialized agents focus on dedicated sub-tasks, including geographic event extraction, temporal element parsing, spatial element localization, and spatial location reasoning. Experiments on both a baseline dataset and a spatial reasoning-enhanced dataset constructed from DUEE demonstrate the method's superior few-shot spatio-temporal parsing capabilities. Specifically, on the baseline dataset, our method attains spatial element parsing performance of F₁@100 m=0.779 and temporal element parsing performance of F_1time=0.856, yielding absolute improvements of 16.3% and 22.1% over the current state-of-the-art baseline, respectively. Then, Ablation studies further validate the effectiveness of the proposed collaborative framework design. Furthermore, a case study analyzing social media data on the “July 20 heavy rainstorm in Zhengzhou” event illustrates the method's effectiveness in capturing the spatio-temporal progression and enhancing the understanding of event evolution. This study introduces a novel multi-agent processing paradigm for geographic event extraction, offering robust support for the construction of refined STKG and the advancement of event-driven geospatial intelligence.

Key words: geographic event, event extraction, spatio-temporal parsing, multi-agent, large language model

中图分类号:

P208

胡鑫, 杨学习, 江一凡, 王宪彬, 丁晨, 谢顾然, 邓敏. 基于多智能体层次化协同的地理事件抽取与时空解析[J]. 测绘学报, 2025, 54(11): 2052-2067.

Xin HU, Xuexi YANG, Yifan JIANG, Xianbin WANG, Chen DING, Guran XIE, Min DENG. Hierarchical multi-agent collaboration for geographic event extraction and spatio-temporal parsing[J]. Acta Geodaetica et Cartographica Sinica, 2025, 54(11): 2052-2067.

图/表 17

图1

表1

图2

图3

图4

图5

表2

表3

表4

图6

表5

表6

图7

图8

图9

图10

图11

参考文献 43

[1]	WORBOYS M. Event-oriented approaches to geographic phenomena[J]. International Journal of Geographical Information Science, 2005, 19(1): 1-28.
[2]	LEE R, SUMIYA K. Measuring geographical regularities of crowd behaviors for Twitter-based geo-social event detection[C]//Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Location Based Social Networks. New York: ACM Press, 2010: 1-10.
[3]	杜云艳, 易嘉伟, 薛存金, 等. 多源地理大数据支撑下的地理事件建模与分析[J]. 地理学报, 2021, 76(11): 2853-2866.
	DU Yunyan, YI Jiawei, XUE Cunjin, et al. Modeling and analysis of geographic events supported by multi-source geographic big data[J]. Acta Geographica Sinica, 2021, 76(11): 2853-2866.
[4]	慎利, 徐柱, 李志林, 等. 从地理信息服务到地理知识服务：基本问题与发展路径[J]. 测绘学报, 2021, 50(9): 1194-1202. DOI: 10.11947/j.AGCS.2021.20210183.
	SHEN Li, XU Zhu, LI Zhilin, et al. From geographic information service to geographic knowledge service: research issues and development roadmap[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50(9): 1194-1202. DOI: 10.11947/j.AGCS.2021.20210183.
[5]	陆锋, 诸云强, 张雪英. 时空知识图谱研究进展与展望[J]. 地球信息科学学报, 2023, 25(6): 1091-1105.
	LU Feng, ZHU Yunqiang, ZHANG Xueying. Spatiotemporal knowledge graph: advances and perspectives[J]. Journal of Geo-information Science, 2023, 25(6): 1091-1105.
[6]	LIU Pengyuan, BILJECKI F. A review of spatially-explicit GeoAI applications in urban geography[J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 112: 102936.
[7]	HU Yingjie. Geo-text data and data-driven geospatial semantics[J]. Geography Compass, 2018, 12(11): e12404.
[8]	刘纪平, 王勇, 胡燕祝, 等. 互联网泛在地理信息感知融合技术综述[J]. 测绘学报, 2022, 51(7): 1618-1628. DOI: 10.11947/j.AGCS.2022.20220102.
	LIU Jiping, WANG Yong, HU Yanzhu, et al. A review of web-based ubiquitous geospatial information discovery and integration technology[J]. Acta Geodaetica et Cartographica Sinica, 2022, 51(7): 1618-1628. DOI: 10.11947/j.AGCS.2022.20220102.
[9]	FENG Yu, HUANG Xiao, SESTER M. Extraction and analysis of natural disaster-related VGI from social media: review, opportunities and challenges[J]. International Journal of Geographical Information Science, 2022, 36(7): 1275-1316.
[10]	余丽, 陆锋, 张恒才. 网络文本蕴涵地理信息抽取：研究进展与展望[J]. 地球信息科学学报, 2015, 17(2): 127-134.
	YU Li, LU Feng, ZHANG Hengcai. Extracting geographic information from web texts: status and development[J]. Journal of Geo-information Science, 2015, 17(2): 127-134.
[11]	QIU Qinjun, XIE Zhong, WU Liang, et al. Automatic spatiotemporal and semantic information extraction from unstructured geoscience reports using text mining techniques[J]. Earth Science Informatics, 2020, 13(4): 1393-1410.
[12]	HOGENBOOM F, FRASINCAR F, KAYMAK U, et al. An overview of event extraction from text[C]//Proceedings of 2011 the DeRiVE Workshop at the International Semantic Web Conference. Aachen: CEUR-WS.org, 2011: 48-57.
[13]	项威, 王邦. 中文事件抽取研究综述[J]. 计算机技术与发展, 2020, 30(2): 1-6.
	XIANG Wei, WANG Bang. Survey of Chinese event extraction research[J]. Computer Technology and Development, 2020, 30(2): 1-6.
[14]	RILOFF E. Automatically constructing a dictionary for information extraction tasks[C]//Proceedings of 1993 National Conference on Artificial Intelligence. Washington, DC: AAAI, 1993.
[15]	GUO Yunhao. A review of event extraction[C]//Proceedings of 2024 International Conference on Electronic Information Technology and Computer Engineering. Haikou: ACM Press, 2024: 1-4.
[16]	LI Qian, PENG Hao, LI Jianxin, et al. A comprehensive survey on schema-based event extraction with deep learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 35(5): 6301-6321.
[17]	CROOKS A, CROITORU A, STEFANIDIS A, et al. Earthquake: Twitter as a distributed sensor system[J]. Transactions in GIS, 2013, 17(1): 124-147.
[18]	ROSSI C, ACERBO F S, YLINEN K, et al. Early detection and information extraction for weather-induced floods using social media streams[J]. International Journal of Disaster Risk Reduction, 2018, 30: 145-157.
[19]	JOSHI A, KARIMI S, SPARKS R, et al. Survey of text-based epidemic intelligence[J]. ACM Computing Surveys, 2020, 52(6): 1-19.
[20]	张雪英, 闾国年, 李伯秋, 等. 基于规则的中文地址要素解析方法[J]. 地球信息科学学报, 2010, 12(1): 9-16.
	ZHNAG Xueying, LÜ Guonian, LI Boqiu, et al. Rule-based approach to semantic resolution of Chinese addresses[J]. Journal of Geo-information Science, 2010, 12(1): 9-16.
[21]	SHAN Shuangli, LI Zhixu, YANG Qiang, et al. Geographical address representation learning for address matching[J]. World Wide Web, 2020, 23(3): 2005-2022.
[22]	LIN Yue, KANG Mengjun, WU Yuyang, et al. A deep learning architecture for semantic address matching[J]. International Journal of Geographical Information Science, 2020, 34(3): 559-576.
[23]	HUANG Jizhou, WANG Haifeng, SUN Yibo, et al. ERNIE-GeoL: a geography-and-language pre-trained model and its applications in Baidu maps[C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Washington, DC: ACM Press, 2022: 3029-3039.
[24]	DING Wentao, CHEN Jianhao, E Longfei, et al. Time expression as update operations: normalizing time expressions via a distantly supervised neural semantic parser[J]. Knowledge-Based Systems, 2023, 278: 110870.
[25]	DONG Xiaofei, ZHANG Xueqiang, BU Weixin, et al. A survey of LLM-based agents: theories, technologies, applications and suggestions[C]//Proceedings of 2024 International Conference on Artificial Intelligence, Internet of Things and Cloud Computing Technology. Wuhan: IEEE, 2024: 407-413.
[26]	WANG Lei, MA Chen, FENG Xueyang, et al. A survey on large language model based autonomous agents[J]. Frontiers of Computer Science, 2024, 18(6): 186345.
[27]	WEI Cheng, ZHANG Yifan, ZHAO Xinru, et al. GeoTool-GPT: a trainable method for facilitating large language models to master GIS tools[J]. International Journal of Geographical Information Science, 2025, 39(4): 707-731.
[28]	吴华意, 赵安琪, 梁健源, 等. 面向地学分析AI建模的地理信息服务层次网络模型[J]. 测绘学报, 2024, 53(11): 2053-2063. DOI: 10.11947/j.AGCS.2024.20240109.
	WU Huayi, ZHAO Anqi, LIANG Jianyuan, et al. Five-layer hierarchical network (5-HiNet) of geospatial information service for AIGC of geographic analysis model[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(11): 2053-2063. DOI: 10.11947/j.AGCS.2024.20240109.
[29]	DANG Pei, ZHU Jun, LI Weilian, et al. A large language model-based agent for wayfinding: simulation of spatial perception and memory[J]. Cartography and Geographic Information Science, 2025, 52(4): 350-369.
[30]	李志林, 徐柱, 慎利, 等. 自主式情境化地图表达：大模型时代的智能化地图制图理论探讨[J]. 测绘学报, 2024, 53(11): 2043-2052. DOI: 10.11947/j.AGCS.2024.20240222.
	LI Zhilin, XU Zhu, SHEN Li, et al. Autonomous situatedness map representation: a theoretical discussion on intelligent cartography in the era of large models[J]. Acta Geodaetica et Cartographica Sinica, 2024, 53(11): 2043-2052. DOI: 10.11947/j.AGCS.2024.20240222.
[31]	SONG Jiayu, ZHANG Yifan, WANG Zhiyun, et al. Large language model agent with VGI data for mapping[J]. Journal of Geodesy and Geoinformation Science, 2025, 8(2): 57-73.
[32]	DONG T, SUBIA-WAUD C, HOU S. Geo-RAG: gaining insights from unstructured geological documents with large language models[C]//Proceedings of the 4th EAGE Digitalization Conference & Exhibition. Paris: European Association of Geoscientists & Engineers, 2024: 1-4.
[33]	WANG Siqin, HU Tao, XIAO Huang, et al. GPT, large language models (LLMs) and generative artificial intelligence (GAI) models in geospatial science: a systematic review[J]. International Journal of Digital Earth, 2024, 17(1): 2353122.
[34]	SUN Chenzhen, LAN Tian, WU Zhiwei, et al. Generative artificial intelligence and its applications in cartography and GIS: an exploratory review[J]. Journal of Geodesy and Geoinformation Science, 2025, 8(2): 74-89.
[35]	LI Xinyi, WANG Sai, ZENG Siqi, et al. A survey on LLM-based multi-agent systems: workflow, infrastructure, and challenges[J]. Vicinagearth, 2024, 1(1): 9.
[36]	YAO Shunyu, ZHAO Jeffrey, YU Dian, et al. React: synergizing reasoning and acting in language models[C]//Proceedings of 2023 International Conference on Learning Representations. Kigali: ICLR, 2023.
[37]	LIU Chengyuan, ZHAO Fubang, KANG Yangyang, et al. RexUIE: a recursive method with explicit schema instructor for universal information extraction[C]//Proceedings of 2023 Conference of Empirical Methods in Natural Language Proceeding. Singapore: ACL, 2023: 15342-15359.
[38]	BI Xiao, CHEN Deli, CHEN Guanting, et al. Deepseek llm: scaling open-source language models with longtermism[EB/OL]. [2025-10-16]. http://arxiv.org/abs/2401.02954.
[39]	王大力, 童晓冲, 孟丽, 等. 文本中空间信息的结构化建模与语义定位[J]. 测绘学报, 2023, 52(8): 1398-1410. DOI: 10.11947/j.AGCS.2023.20220066.
	WANG Dali, TONG Xiaochong, MENG Li, et al. Structural modeling of spatial information in texts and semantic localization[J]. Acta Geodaetica et Cartographica Sinica, 2023, 52(8): 1398-1410. DOI: 10.11947/j.AGCS.2023.20220066.
[40]	LI Xinyu, LI Fayuan, PAN Lu, et al. DuEE: a large-scale dataset for Chinese event extraction in real-world scenarios[C]//Proceedings of 2020 Natural Language Processing and Chinese Computing Conference. Zhengzhou: Springer, 2020: 534-545.
[41]	LU Yaojie, LIU Qing, DAI Dai, et al. Unified structure generation for universal information extraction[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. Dublin: ACL, 2022: 5755-5772.
[42]	ABOUDZADEH N, HAJIPOOR SHOSHTARI A, HASHEMNIA S. Crisis management: planning for the inevitable[J]. Management Science Letters, 2014, 4(6): 1191-1196.
[43]	河南郑州“7·20”特大暴雨灾害调查报告[EB/OL]. [2025-04-20]. https://www.mem.gov.cn/gk/sgcc/tbzdsgdcbg/202201/P020220121639049697767.pdf.
	Investigation report on the July 20 catastrophic rainstorm disaster in Zhengzhou, Henan province[EB/OL]. [2025-04-20]. https://www.mem.gov.cn/gk/sgcc/tbzdsgdcbg/202201/P020220121639049697767.pdf.

类别	数据项	格式示意（<…>表示占位符）	示例
输入	原始文本	无格式要求	1月22日6时许，云南镇雄发生滑坡，据悉位置在凉水村雷家院子西南方向约100米处，已造成18户房屋被掩埋、44人遇难
输入	事件本体	{"<事件类型>（触发词）"：{"<论元类型1>"：None，"<论元类型2>"：None，…}	{"灾害（触发词）"：{"位置"：None，"时间"：None，"死亡人数"：None，"受灾情况"：None}}
	地理事件抽取结果	{"<事件类型>（触发词）"："<触发词文本>"，"<语义论元1>"："<论元文本>"，"<语义论元2>"："<论元文本>"，…"<时间论元>"："<时间描述文本>"，"<空间论元>"："<空间位置描述文本>"}	{"灾害（触发词）"："滑坡"，"死亡人数"："44人"，"受灾情况"："18户房屋被掩埋"，"时间"："1月22日6时许"，"位置"："云南镇雄凉水村雷家院子西南方向约100米处"}
输出	时间要素解析结果	{"标准时间"："<时间点/时间段文本^①>"}	{"标准时间"："YYYY/01/22-06"}
输出	空间要素解析结果	{"标准位置"："<位置描述文本^②>"，"中心点坐标"："<（lon，lat）^③>"，"AOI信息"："<GeoJSON文件存储路径>"}	{"标准位置"："云南省昭通市镇雄县塘房镇凉水村雷家院子西南方向约100米处"，"中心点坐标"："（105.0135，27.4775）"，"AOI信息"："/example/path.json"}

方法	模型版本	超参数	提示策略
PTMs Pipeline	uie/rexuie/JioNLP	N/A	N/A
Single Agent MAGESTP	qwen-max-20250125/deepseek-v3-20250324	temperature=0.2，top_p=1，max_token=100 000	将完整任务在一个提示中描述为不同智能体设计专用提示

模型方法	空间要素解析指标			时间要素解析指标
模型方法	P@100 m	R@100 m	F₁@100 m	P_time	R_time	F_1time
PTMs Pipeline（UIE）	0.554	0.593	0.573	0.648	0.592	0.618
PTMs Pipeline（RexUIE）	0.564	0.656	0.606	0.702	0.594	0.643
Single Agent（Qwen-max）	0.638	0.705	0.670	0.698	0.703	0.701
Single Agent（DeepSeek-V3）	0.632	0.696	0.663	0.683	0.708	0.695
MAGESTP（Qwen-max）	0.738	0.731	0.734	0.769	0.786	0.777
MAGESTP（DeepSeek-V3）	0.804	0.756	0.779	0.839	0.873	0.856

模型方法	空间要素解析指标
模型方法	P@100 m	R@100 m	F₁@100 m
PTMs Pipeline（UIE）	0.459	0.457	0.458
PTMs Pipeline（RexUIE）	0.470	0.506	0.487
Single Agent（Qwen-max）	0.517	0.544	0.530
Single Agent（DeepSeek-V3）	0.509	0.537	0.523
MAGESTP（Qwen-max）	0.644	0.635	0.639
MAGESTP（DeepSeek-V3）	0.729	0.694	0.711

模型方法	空间要素解析指标			时间要素解析指标
模型方法	P@100 m	R@100 m	F₁@100 m	P_time	R_time	F_1time
MAGESTP	0.729	0.694	0.711	0.851	0.884	0.867
-w/o Ensemble	0.718	0.693	0.705	0.834	0.868	0.851
-w/o SA₄	0.608	0.598	0.603	0.841	0.889	0.865
-w/o Ensemble & SA₄	0.608	0.592	0.600	0.844	0.877	0.860

基于多智能体层次化协同的地理事件抽取与时空解析

Hierarchical multi-agent collaboration for geographic event extraction and spatio-temporal parsing

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 17

参考文献 43

相关文章 3

编辑推荐

Metrics

本文评价

抽取字段	含义
事件触发词	文本中用于指示暴雨或洪水灾害发生的关键短语
时间（事件论元）	文本中事件发生的具体时间
地点（事件论元）	文本中事件发生的具体地理位置
事件摘要（事件论元）	对事件主要信息进行简明概括
事件主题（事件论元）	反映事件的内容焦点，类别包括：灾害预警提示、灾情实况信息、人员伤亡情况、设施损毁情况、官方应急救援、社会力量援助、求助信息发布、灾后恢复重建以及其他

[1]	吴华意, 沈张骁, 侯树洋, 梁健源, 赵安琪, 矫皓月, 桂志鹏, 关雪峰. 大语言模型驱动的GIS分析：方法、应用与展望[J]. 测绘学报, 2025, 54(4): 621-635.
[2]	汤圣君, 杜思齐, 王伟玺, 郭仁忠. 面向室内空间智能的三维场景图表达与应用[J]. 测绘学报, 2024, 53(7): 1355-1370.
[3]	刘纪平,栗斌. 一种本体驱动的地理空间事件相关信息自动检索方法[J]. 测绘学报, 2011, 40(4): 0-441.