热点发现及文本倾向性分析技术研究

Research of Hot-topic Finding and Text Orientation Analysis

为了有效地对不断涌现的海量互联网信息进行采集、分析、检索,基于网络舆情分析涉及到的热点发现与文本倾向性分析两个关键技术,改进了相似主题检测和基于语气标注方法的文本倾向性分析算法。实验结果表明,直接通过关键词匹配法,误差较大;基于关联规则的相似主题检测,可以明显提高检测精度。同时,无论是对正面文档还是对负面文档,经过改进的基于语气标注方法都明显提高了查全率和查准率。

In order to effectively collect,analyze and retrieve the emerging mass of Internet network information,based on two key technologies concerning online public opinion analysis（hot-topic finding and text orientation analysis）,similar topic detection and tone marking-based text orientation analysis algorithm are improved.Experimental results show that a higher error occurs by directly using keyword matching while detection accuracy can get an obvious improvement through similar topic detection based association rules.Moreover,it concludes that the modified tone marking-based method is notably effective to improve both recall and precision whether it is applied in detecting positive or negative documents.