档案管理中文本数据的增量多模态聚类方法  

Incremental multi-modal clustering methods fortext data in archives administration

作　　者：刘丽华[1] LIU Lihua(Archives,Inner Mongolia University of Finance and Economics,Hohhot 010010,P.R.China)

机构地区：[1]内蒙古财经大学档案馆,呼和浩特010010

出　　处：《重庆大学学报》2022年第5期147-156,共10页Journal of Chongqing University

基　　金：内蒙古自治区教育科学“十三五”规划2019年度课题(NGJGH2019360);内蒙古财经大学2019年校级教育教学课题(JXYB1924)。

摘　　要：随着现代档案管理数据量的不断增长,有效地对档案文本进行聚类划分能够提升档案分类和检索的效率。文中提出2种增量多模态文本数据聚类方法,通过对文本内容进行多视角分析,融合挖掘文本的潜在主题特征,提升文本聚类的准确性。此外,设计文本聚类多模态增量学习模型,提升海量、动态文本划分的效率。在文本数据集上的实验结果表明,文中提出的增量多模态文本聚类方法优于单模态和多模态聚类算法,能够对文本数据进行有效划分。With the continuous growth of modern archive management data,the effective clustering of archive text can significantly improve the efficiency of archive classification and retrieval.This paper proposes two incremental multi-modal text data clustering methods.By multi-perspective analysis of the text content,the potential topic features of texts are integrated to improve the accuracy of text clustering.In addition,the corresponding incremental multi-modal feature learning models for text clustering are designed to improve the efficiency of massive and dynamic text partition.Experimental results on real-world text data sets show that the proposed incremental multimodal text clustering methods outperform the compared stated-of-the-art methods,being able to effectively classify text data.

关 键 词：文本数据 多模态聚类 增量特征学习 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]