基于关联分析的偏离主题探测

Off-Topic Detection Using Association Analysis

互联网中的信息良莠不齐,因此必须对信息的发布、传播和访问进行有效的监控。离题文档检测指通过主题相关性来界定访问文档的合法性。超团模式是一种附加了整体相似度约束的特殊频繁项集。利用超团这种特性,提出了基于关联分析的离题文档检测方法,并介绍了原型系统的实现及应用。

Due to the openness,not all information on the Internet is good,so we must effectively monitor the publish,dissemination and access of harmful information.Off-topic detection refers to detecting illegitimate access according to topic relevance to users＇predefined area of interest.Hyperclique patterns are a special type of frequent itemsets that are constrained by group similarity.Along this line,this paper proposes an off-topic detection method using association analysis.Finally,we briefly introduce its application to a prototype system.