基于文件标题特征的网络视频去重研究  

Research on Duplicate Removal for Network Video Based on File Title Features

作　　者：郭牧怡[1,2] 刘萍[1] 谭建龙[1] 郭莉[1] 

机构地区：[1]中国科学院计算技术研究所,北京100190 [2]中国科学院研究生院,北京100049

出　　处：《计算机工程》2010年第9期227-229,237,共4页Computer Engineering

基　　金：国家"973"计划基金资助项目(2007CB311103);国家"863"计划基金资助重点项目(2007AA010501)

摘　　要：提出一种针对剧集类和电影类网络视频的去重技术,根据剧集类和电影类网络视频标题的显著特征,采用精确串和正则表达式匹配算法自动提取视频名称,形成以正则表达式表示的视频扩展名,当新视频文件到来时,再次使用正则表达式匹配算法,判断其是否出现过,从而实现网络视频的去重。实验结果表明,该方法的查准率可达97.30%,查全率可达93.63%。A method for removing duplicates among teleplays and movies is presented.Titles of these network videos usually presents distinct features which make it possible to extract videos’ names automatically by means of exact pattern matching and regular expression matching algorithms.The extracted names is processed further into regular expressions against which up-coming video titles are matched so as to accomplish the goal of removing duplicates.Experimental results show this method achieves an accuracy ratio of 97.30% and a recall ratio of 93.63%.

关 键 词：正则表达式 精确串匹配算法 网络视频 去重 

分 类 号：N945[自然科学总论—系统科学]