基于统计-规则方法的网页层次分类技术研究.pdfVIP

《现代图书情报技术 2∞7 年第 8 期 知识组织与知识管理 总第 154 期 基于统计…规则方法的网页层次分类技术研究 谭金波 杨晓江李艺 (山东师范大学教育技术系 济南 250014) (南京师范大学教育技术系 南京 210097) [摘要] 基于统计的自动分类是网页层次分类中常用的技术，但具有不足之处，主要表现为当于英之间出 现严重的特征交叉现象时，分类精确率将大大下降。而网页层次分类的本质决定了同一大英下的子英存在许 多相同的特征。针对这一局限性，结合蕃于规则的自动分类技术的优点，提出一种篡于统计咀规则方法的网页 层次分类技术。实验表明，基于统计-规则方法的问页层次分类技术能够获得比较理想的分类效果。 [提键词] 规则分3传统计分类网页层lX分类基于统汁规则的分类 [分提号 1 G354.4 Study of Statistics - rule ?ased Hierarchical Web Page Classification Tan Jinbo (Educαtional TechMlogy DepaTtment , Shαndong Normal University ,Jinan 250014 , China) Yang Xiaojiang Li Yi (Educαtional TechMlogy DepαTtment ,Nanjing Normal University ， 胁时ing 21 ω97 , China) [ Abstract] Statistics - h叫 classífication methods are common -叫 in hiera时lical Web classification. How? ever, classification precision of statistics - based methods often drops when categories are veηsimilar ωeach other be? cau串串 of feature overlapping. Due to the nature of hie阳chical Web classification , categories sharing the same 阳rent (e. 岛， Ieaf categories in 由e hierarchy) 缸e often very similar ωeach other. To impro冈山e classification precision , the p衍 per pro严剧由 ωωe rule - based classification me出ods on top of statistics - based methods in hierarchical Web classifi? cation. Experiments show that our metl时s perform wel1 on our education Web collections. [Keywor也] Statistics - based c1assification Rule - based classification Hierarchi叫 Web c1assification Statistics - rule based classifi