基于多特征融合的网页正文信息抽取  

EXTRACTING WEBPAGES TEXT INFORMATION BASED ON HETEROGENEOUS FEATURES FUSION

作　　者：刘利[1] 戴齐[1] 尹红风[1] 贾真[1] 胡万亭[1] 

机构地区：[1]西南交通大学信息科学与技术学院,思维与智慧研究所,四川成都610031

出　　处：《计算机应用与软件》2014年第7期47-49,77,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61152001;61170111);中国科学院自动化研究所复杂系统管理与控制重点实验室开放课题(20110102)

摘　　要：当今主流网页分为单正文体网页和多正文体网页。这些网页的正文信息都具有多个正文特征。想要准确定位正文信息所在位置,可以从其所具有的多个特征和网页设计者的设计习惯着手。鉴于此,融合这些特征提出一种基于多特征融合的网页正文信息抽取方法。实验结果表明,该方法对单正文体网页和多正文体网页的正文抽取具有较高的准确率和通用性,很好地适应了风格多样的网页。Nowadays, the mainstream webpages are divided into single text body webpages and multiple text body webpages. These webpages text information all have the heterogeneous text features. In order to accurately position the location of text information, one can commence from the heterogeneous features they possess and the design habits of the web designers. Therefore, in this paper, we fuse these features and present a webpages text information extraction method which is based on heterogeneous features fusion. Experimental results show that the method has higher accuracy rate and universality for extracting the texts from single test body webpages and multiple text body webpages, fits well the webpages with a variety of styles.

关 键 词：单正文体 多正文体 多特征 信息抽取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]