机器学习在档案管理中的应用：进展与挑战

The Application of Machine Learning in Archives Management:Progress and Challenges

自2013年起,陆续有学者和档案馆开展了机器学习在档案管理中的应用研究,涉及手写档案识别、档案分类、档案鉴定、利用控制和利用服务等诸多业务。本文基于机器学习的技术逻辑和应用逻辑,结合档案及其管理工作的特点,分析得出机器学习在档案管理领域的应用逻辑框架,结合现有的应用案例,总结了机器学习在档案管理领域的应用进展,并分析了深化应用面临的挑战。研究发现,现阶段机器学习在档案管理领域的应用以实验性研究为主,在部分重要、棘手的档案管理业务展开初步探索,主要以文本数据和图片数据为训练数据集,以有监督的分类任务作为主要的任务形式。机器学习在档案管理领域中深化应用可能面临来自档案管理、档案数据、机器信任、法律环境和专业能力的挑战。

In the field of archial science,researchers and archives have already put machine learning into application and practice since 2013 mainly including handwriting archives recognizing,archives classifying,archives appraisal,using control and using service.We put forward the framework of how machine learning applies in archives management from the technical logic and applied logic of machine learning,and analyze the progress of the machine learning application in archives management based on some representative cases.From these cases,we find 4 characteristics at present stage:nearly all experimental research;tentative exploration in some importance or intractable problem;most of the datasets are text or images;supervised classification is the major form of these applications.Furthermore,based on the analyzation we put forward 5 challenges that may arise with the deepening of machine learning application in archives management including:the challenge of archive management,the challenge of archives data,the challenge of machine trust,the challenge of legal environment and the challenge of professional ability.