全球战略性矿产资源信息感知技术方法构建及模型研究  

Research on the construction of information perception technology methods and models for global strategic mineral resources

作　　者：肖飞 曾建鹰 李政 陈从喜 郭文华 张玉韩 朱力维 仇巍巍 XIAO Fei;ZENG Jianying;LI Zheng;CHEN Congxi;GUO Wenhua;ZHANG Yuhan;ZHU Liwei;QIU Weiwei(Information Center of Ministry of Natural Resources,Beijing 100036,China;Technology Innovation Center for Territorial&Spatial Big Data,Ministry of Natural Resources,Beijing 100830,China;School of Information Engineering,China University of Geosciences(Beijing),Beijing 100083,China)

机构地区：[1]自然资源部信息中心,北京100036 [2]自然资源部国土空间大数据工程技术创新中心,北京100830 [3]中国地质大学(北京)信息工程学院,北京100083

出　　处：《中国矿业》2025年第2期48-56,2,共10页China Mining Magazine

基　　金：国家重点研发计划项目课题“全球战略性矿产大数据平台和预警与决策支持技术”资助(编号:2021YFC2901801);国家重点研发计划项目“国家国土空间规划实施监测系统研发与示范”资助(编号:2023YFC3804005)。

摘　　要：收集和分析全球战略性矿产资源信息情报对于建立战略性矿产监测预警机制至关重要。得益于互联网的飞速发展,各国政府、矿业企业和研究机构等通过互联网公开发布了海量的矿产资源信息,但这些信息具备鲜明的大数据特征,使得感知分析面临巨大挑战。本文研究了面向矿产资源信息感知的大数据技术方法和信息感知模型,构建了“采集—处理—存储”一体化的全球战略性矿产资源信息感知大数据技术框架,解决了全球战略性矿产资源信息实时监测和分析的数据采集和信息提取问题;面向全球战略性矿产资源全生命周期重要信息节点,针对性地提出了实体识别、关系抽取、事件抽取和文本分类四项全球战略性矿产资源信息感知任务,构建了万条规模的矿产资源信息感知数据集,建立了基于BERT的全球战略性矿产资源信息感知模型,解决了非结构化矿产资源信息语义内容的结构化提取问题。实验表明,模型在各项感知任务的精准率和召回率指标均达到0.75以上,且各项任务预测的平均值达到0.80以上,证明模型具有较强的泛化能力,能够有效理解矿产资源领域专业术语要素,精准捕捉要素关联关系,具有较高的实用价值。本文提出了一套自动化的全球战略性矿产资源信息感知技术方法和模型,可为全球战略性矿产资源安全预警提供信息感知技术支撑。The collection and analysis of global strategic mineral resource intelligence is crucial for establishing monitoring and early warning mechanisms.Benefiting from rapid Internet development,governments,mining enterprises,and research institutions have released massive mineral resource information online publicly.However,such information exhibits distinct big data characteristics that pose significant challenges for perception analysis.This paper investigates big data methodologies and information perception models for mineral resource intelligence analysis,constructing an integrated“collection-processing-storage”technical framework for global strategic mineral resource information perception.This framework resolves critical issues in real-time monitoring and data acquisition for global mineral resource analysis.Focusing on key information nodes throughout the lifecycle of strategic mineral resources,it specifically propose four perception tasks:entity recognition,relation extraction,event extraction,and text classification.A 10000-entry mineral resource perception dataset is constructed,and a BERT-based information perception model is developed to address the structured extraction of semantic content from unstructured mineral resource data.Experimental results demonstrate that the model achieves precision and recall rates exceeding 0.75 across all perception tasks,with an average score across tasks surpassing 0.80.This confirms the model’s strong generalization capability in comprehending domain-specific terminology and accurately capturing elemental correlations,indicating high practical value.This paper presents an automated information perception technical methodology and model for global strategic mineral resources,providing critical information sensing support for early warning systems of strategic mineral resource security.

关 键 词：战略性矿产资源 信息感知 大数据 感知模型 预训练 BERT模型 

分 类 号：TD-0[矿业工程]