基于ODP的上下文主题描述方法  

A Novel Context-Sensitive Topic Description Method Based on ODP

作　　者：吴麒[1] 陈兴蜀[1] 朱锴[1] 王春晖[1] 

机构地区：[1]四川大学计算机学院网络与可信计算研究所,四川成都610065

出　　处：《电子学报》2012年第11期2320-2323,共4页Acta Electronica Sinica

基　　金：国家科技支撑计划课题(No.2012BAH18B05);新媒体资源管理关键技术研究及原理系统

摘　　要：针对以往主题描述方法未充分考虑主题上下文的问题,提出了基于ODP(开放式分类目录)的上下文主题描述方法.使用新的特征选择算法对主题特征进行了确定,并使用分类主题树的上下文对主题描述方法进行优化以提高主题爬行的性能.实验表明,该特征选择算法能够有效地提取出主题特征,并在保证正确率的基础上尽量减少特征维数以提高计算效率.同时,该主题描述算法充分考虑了主题上下文关系,且无论是在准确性还是在信息量总和上都有良好的性能.It becomes an increasing important task to discover desirable information from tremendous web resources,and therefore the more accurately the focused crawler describes the users＇ interested topics,the more useful the information obtained by it will be.However,the neglect of context leads traditional approaches to the failure of achieving this goal.In order to overcome this shortcoming,a new context-sensitive topic description method based on ODP（Open Directory Project） is proposed in this paper.First of all,the best topic feature subset is determined by a new feature selection algorithm.And then,we optimize the topic description method to improve the performance by utilizing topic context.The experimental results show that this new approach extracts the features effectively with a better performance in both the precision and the sum of information while minimizing its dimension size significantly.

关 键 词：主题爬行 下文相关 特征选择 主题描述 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]