基于信息粒度的主题相似性信息检索

Topic similarity information retrieval based on information granularity

主题相似性计算是信息检索、文本分类等领域内的一个研究热点。基于信息粒度原理,内容主题识别和事件主题识别是在不同粒度世界进行的识别计算。本研究将传统的内容主题识别算法与基于TDT的事件主题识别算法相结合,提出1种新的主题相似性计算模型,即先进行内容主题识别,再进行事件主题识别。试验结果表明:该模型具有良好的分类效果。

Topic similarity computation model is a research hotspot in fields of information retrieval and text classification.From the view of information granularity,it is very clear that similarity computations of topic identification are under different granularity.According to this principal,this paper integrates traditional content topic identification with event topic identification,and presents a new topic similarity computation model,i.e.,first carrying out content identification,secondly event identification.The experiments show that this model can obtain satisfactory classification result.