泛娱乐情报主题的感知研究  

Research on awareness of pan-entertainment information subjects

作　　者：于汝意 刘秀磊[2] 刘旭红[1] 张良[2] 王延飞[3] YU Ruyi;LIU Xiulei;LIU Xuhong;ZHANG Liang;WANG Yanfei(Key Laboratory of Internet Culture and Digital Dissemination Research,Beijing Information Science&Technology University,Beijing 100101,China;Laboratory of Data Science and Intelligence Analysis Research,Beijing Information Science&Technology University,Beijing 100101,China;Department of InformationManagement,Peking University,Beijing 100871,China)

机构地区：[1]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100101 [2]北京信息科技大学数据与科学情报分析实验室,北京100101 [3]北京大学信息管理系,北京100871

出　　处：《北京信息科技大学学报（自然科学版）》2020年第2期58-61,共4页Journal of Beijing Information Science and Technology University

基　　金：国家重点研发计划资助(2017YFB1400402);北京市教育委员会科技计划面上项目(KM201811232018);网络文化与数字传播北京市重点实验室开放课题资助(ICDDXN006);北京信息科技大学“勤信人才”培育计划资助;北大方正集团有限公司数字出版技术国家重点实验室专项课题资助。

摘　　要：针对泛娱乐情报存在繁杂、异构、信息量较少的特点以及传统主题发现方法感知结果准确率低等问题,提出了一种基于潜在语义索引结合常识知识库的泛娱乐情报主题感知方法。根据泛娱乐情报的语料特征,首先在预处理阶段采用HowNet知识库进行语义去重、同义词合并,形成文本向量作为潜在语义索引模型的输入;然后通过潜在语义索引模型,对泛娱乐情报进行主题感知。该方法是主题模型在泛娱乐领域情报分析中的一次尝试,具有一定的实践意义。Aiming at the problems of pan-entertainment intelligence being complicated,heterogeneous,and less informative,and the low accuracy rate of traditional subject discovery methods,a pan-entertainment information subject awareness method is proposed based on latent semantic index and common knowledge base.On the basis of the characteristics of pan-entertainment information in the corpus,firstly,the HowNet knowledge base is used to form a text vector as the input of the latent semantic indexing model for semantic deduplication and synonym merging in the preprocessing stage;then the subject awareness of pan-entertainment information is performed through the latent semantic indexing model.This method is a valuable attempt to analyze information in the pan-entertainment field,which has certain practical significance.

关 键 词：情报感知 潜在语义索引 HOWNET 文本挖掘 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]