基于粒度商空间下的话题识别与跟踪研究-《计算机技术与发展》

文章信息/Info

Title:: Research on Topic Recognition and Tracking Based on Granular Quotient Space

Author(s):: MAO Jian-jing; ZHANG Jun-jun; School of Information Technology,Zhengzhou University of Industrial Technology,Zhengzhou 451150,China

Keywords:: compatible quotient space; granularity; topic identification; public opinion

摘要:: 文中旨在对自然语言所形成的信息流进行话题识别与跟踪,其目的主要针对网络舆论中出现的新话题进行识别, 并实现对已有话题的跟踪研究。基于相容商空间粒度下软聚类算法,实现对话题的识别与跟踪,是舆情分析的关键技术。话题识别与跟踪采用软聚类算法,根据相容关系原理,计算距离函数,使话题呈现一定的层次结构,再利用相容隶属函数实现对边界文本的话题确认,形成注明标注信息的语料。同时,结合基于 Ontology 情感分类法,计算与情感词汇中的语义相似度,统计目标情感词汇的倾向性权重,建立基于粒度商空间下的话题识别与跟踪模型,有效地促进话题倾向性的研究,最终实现对网络舆情话题的识别与跟踪,为相关部门监管网络舆情、掌握舆论方向提供指导。

Abstract:: This topic aims to identify and track the information flow formed by natural language. Its purpose is to identify the new topics in the network public opinion and to realize the tracking research of the existing topics. The soft clustering algorithm based on compatible quotient space granularity realizes the recognition and tracking of topics,which is the key technology of public opinion analysis. The topic identification and tracking adopts the soft clustering algorithm. According to the principle of compatible relationship,the distance function is calculated to make the topic present a certain hierarchical structure. Then the compatible membership function is used to confirm the topic of the boundary text,and the corpus indicating the annotation information is formed. At the same time,based on Ontology sentiment classification,the semantic similarity in emotional vocabulary is calculated,the tendency weight of target emotional vocabulary is calculated,and the topic recognition and tracking model based on granular quotient space is established to effectively promote the research of topic orientation. Finally,the identification and tracking of online public opinion topics is realized to provide guidance for relevant departments to supervise network public opinion and master the direction of public opinion.