基于切平面的主题提取算法

New cut-plane based algorithm for topic distillation

从语义相关性角度分析超链归纳主题搜索(HITS)算法,发现其产生主题漂移的原因在于页面被投影到错误的语义基上,因此引入局部密集因子LDF(Local Density Factor)的概念。为了解决Web内容的重叠性,基于切平面的概念提出了一种新的主题提取算法(CPTDA)。CPTDA不但可以发现用户最感兴趣的主题页面集合,还可以发现与查询相关的其他页面集合。在10个查询上的实验结果表明,与HITS算法相比,CPTDA算法不仅可以减少30%-52%的主题漂移率,而且可以发现与查询相关的多个主题。

To interpret the procedure of hypertext induced topic search based on a semantic relation model,the reason about the topic drift of HITS has been found that Web pages are projected to a wrong latent semantic basis.A new concept LDF（Local Density Factor） has been introduced and based on cut-plane a new topic distillation algorithm CPTDA（Cut-Plane based Topic Distillation Algorithm） has been presented to improve the quality of topic distillation.CPTDA has been applied not only to avoid the topic drift,but also to explore relative topics of user query.The experimental results on 10 queries show that CPTDA reduces topic drift rate by 30% to 52% compared to that of HITS,and discovers several relative topics to queries that have multiple meanings.