军事知识服务平台
基于隐私计算的档案开放智能审核多主体数据协同模式与方法研究
Models and Methods for Multi-Stakeholder Data Collaboration in Privacy-Preserving Computing-Enabled Intelligent Archival Opening Appraisal
摘要 档案开放智能审核在跨部门、多主体协作场景中的需求日益突出,智能化审核模型的性能依赖于高质量、多样化的数据集.然而,传统数据集中的共享模式在数据安全共享与高效高质量协同等方面存在局限性.针对这些问题,探讨隐私计算技术在档案开放智能审核多主体协同中的应用价值,聚焦档案开放智能审核场景,识别数据协同过程中的多方主体角色,构建"语料-模型-知识"三层级协同体系,形成档案开放智能审核"二任务、三阶段"数据协同方法框架,以数据标注、模型训练与知识应用阶段的逐步衔接与融合,实现数据识别与开放审核任务的智能、高效开展,从而应对数据孤岛、重复建设与隐私风险等实际难题,为我国档案开放智能审核的规模化推进与实践落地提供理论依据与技术参考.
Abstract:
The growing need to appraise archives intelligently in cross-departmental,multi-stakeholder environments demands high-performing models,which require large,diverse,and trustworthy datasets.However,traditional centralized data-sharing approaches fall short in secure data sharing and efficient,high-quality collaboration.In response to these issues,this study explores the application value of privacy-preserving computing in multi-stakeholder collaboration for intelligent archival opening appraisal.Focusing on the archival opening appraisal scenario,it first delineates the roles of all participating actors,then introduces a three-tier"corpus-model-knowledge"collaborative architecture and establishes a"two-task,three-phase"data-collaboration framework,which links and integrates data annotation,model training,and knowledge application to support the dual tasks of sensitive-content identification and openness appraisal.The framework overcomes data silos,duplication of effort,and privacy-leakage risks,offering both theoretical foundations and practical technical guidance for the large-scale implementation and practical application of intelligent archival opening appraisal in China.
刊 名 档案学通讯 ISTICPKU
年,卷(期): 2026, (3)
分类号: G273.5
机标分类号: TP391 F273.1 G64
在线出版日期: 2026年6月17日
基金项目: 2025-Z-007:国家档案局重点科技项目,