Web检索结果快速聚类方法的研究与实现  

Research and realization about fast clustering method of web search engine results

作　　者：王志梅[1] 张俊林[2] 李秋山[1] 

机构地区：[1]燕山大学 信息工程学院 河北 秦皇岛 066004 [2]中国科学院 软件研究所中文信息中心

出　　处：《计算机工程与设计》2004年第12期2231-2233,2290,共4页Computer Engineering and Design

摘　　要：为了帮助Web用户从搜索引擎所返回的大量文档片断中筛选出自己所需要的文档,在对聚类过程研究分析的基础上给出了一种Web检索结果快速聚类方法。它通过分析聚类过程,从建立索引模型、相似性的计算到聚类结果的形成等环节,都做了分析和简化,并利用检索结果的标题、Url以及文档片断3部分所含信息计算返回结果之间的相似度,将首先返回的部分检索结果利用无向图映射法进行部分聚类后,将其余返回结果分配到与之最相近的集簇中最终形成聚类结果。该方法实现简单。实验证明该方法响应速度快,聚类相关性较高,空间占用少。To help web users to choose the wanted documents from large number of snippets returned by search engine, a fast clustering method is provided based on snippets after researching and analyzing the clustering progress. From the building index model, calculating the similarities to clustering the snippets, they are all analysed and simplified. The similarity is calculated by the content of snippets' title, url and abstract. Snippets and their similaryties are mapped to an undirected graph, the part of the first returned snippets and their similarities form the clusters relying on the graph at first, and then the others are assigned to the closest cluster. Experiment shows that the method is easy to compliment and it has a good performance on running-speed, clustering quality and occupated-space.

关 键 词：聚类方法 文档 WEB检索 相似度 搜索引擎 快速 集簇 检索结果 标题 环节 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP311[自动化与计算机技术—计算机科学与技术]