【摘要】 传统的主题提取算法存在一些已知的问题。为了更好地满足SEWM-2004中文W eb检索测评中的主题提取任务要求,分析经典的基于超链接分析的主题搜索(hyperlink-induced top ic search,H ITS)等算法,提出了一种以站点作为查询的资源单位,并结合内容分析的主题提取算法CW T 100G上的超链接分析(hyperlink ana lys is w ith inCW T 100G,HAC)。HAC算法首先根据网页的URL将网页按站点分组;然后在每个站内结合网页内容和站内链接关系来计算网页的权值,从而找出站内的Hub网页;接着再分析站间的链接关系进一步计算各网页的最终权值,从而找出站间的Hub网页。两组对比的实验结果表明,HAC算法能找到切合主题的更大的Hub站点。更多还原

【Abstract】 The traditional algorithms of topic distillation have some known problems, and they can’t meet the requirements of the topic distillation task in the SEWM-2004 Chinese Web Search contest. Based on the analysis of the hyperlink-induced topic search HITS, this paper presents an improved algorithm named hyperlink analysis within CWT100G HAC, which connect content analysis with hyperlink analysis and focus on the site resource. The HAC algorithm groups the pages by site through the URL pattern matching. Then model the inner hyperlinks structure into a graph, together with page content analysis, to calculate the iterative Hub/Authority value inside every site. Afterwards, analyze the outer hyperlinks between sites, and calculate the more accurate Hub/Authority value. The two contrastive experiment results show that the HAC algorithm can find out the more suitable Hub sites for a given query.更多还原