基于图模型的关键词挖掘方法  

Keywords Mining Method Based on Graph Model

作　　者：翟周伟[1] 刘刚[1] 吕玉琴[1] 

机构地区：[1]北京邮电大学电子工程学院,北京100876

出　　处：《软件》2012年第8期9-13,共5页Software

基　　金：国家自然科学基金(60871082)

摘　　要：文章提出一种基于图模型的关键词挖掘方法,应用K最邻近耦合图构造文档的图模型,将文档映射为一个语义结构图,然后结合聚类系数变化量,平均路径变化量,TF-IDF以及区域位置因子来衡量词语节点的重要性,根据重要性得分选择候选关键词集,最后根据短语合并规则形成最终的关键词,实验结果表明,该方法相比于TF-IDF和小世界特征方法性能有所提高。In this paper, a new algorithm is proposed for keywords mining by graph model.Using k-nearest neighbor couple graph model,a document is mapped to a semantic structure graph,then combined with four variables,clustering coefficient increment,average path length increment,TF-IDF and regional location factor, to measure the importance of word nodes, according to the importance score selecte the candidate words set,finally, form the document words by merger rules.Experimental results show that the performance is more efficient than TF-IDF method and Small-World mode method.

关 键 词：计算机应用 中文信息处理 关键词挖掘 复杂网络 综合测度 

分 类 号：P315.69[天文地球—地震学]