智慧军营网上党校中智能化文本分类应用设计  

Design for Intelligent Text Classification Application in Online Party School of Smart Barracks

作　　者：赵艳婷 梅源 苏延庆 宫明煜 林学[1] ZHAO Yanting;MEI Yuan;SU Yanqing;GONG Mingyu;LIN Xue(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210023,China;Unit 32065 of PLA,Shenyang 110161,China)

机构地区：[1]中国电子科技集团公司第二十八研究所,南京210023 [2]解放军32065部队,沈阳110161

出　　处：《指挥信息系统与技术》2023年第6期71-77,共7页Command Information System and Technology

摘　　要：针对智慧军营系统网上党校功能板块的分类学习和推送需求,分析其海量学习任务、新闻通知和组织管理等文档特性,提出了一种高效可行的智能化数据分析设计方案。首先,分析了网上党校各功能模块文本数据的相似性;其次,研究了系统十大建设板块中文本主题划分的聚类融合、文本知识特征抽取及递推关系构建3类关键技术;然后,提出了一种基于党政知识块摘要和深度学习模型的文档分类算法;最后,利用专家方阵、党建研究等板块的文档数据,验证了该算法的有效性,表明该设计方案可通过智能辅助系统为党政工作者提供决策支持,从而实现提高办事质量和效率的目标。Aiming at the requirements of classification learning and classification pushing in the functional block named “online party school” of the smart barracks system,properties of mass learning tasks,news notification and organization management documents are analyzed,and an efficient and feasible intelligent data analysis scheme is proposed.Firstly,the text data similarity of each functional module of the “online party school” is analyzed.Secondly,three kinds of key technology including flustering fusion of text theme partitioning,feature extraction of text knowledge and construction of recursive relation in ten developing blocks of the system are studied.Then,a document classification algorithm of abstracts and deep learning model based on the party and government knowledge block is presented.Finally,by using texts in the blocks named “expert array” and “party building research”,etc.,the validity of the algorithm is verified.It is shown that the design scheme can provide decision support for the party and government workers with the intelligent auxiliary system.Thus,the goal of improving the service quality and efficiency can be realized.

关 键 词：文本分类 知识抽取 聚类融合 玻尔兹曼机 预测推送 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]