Re-ranking search results using network analysis a case study with google: a case study with Google
Re-ranking search results using network analysis a case study with google: a case study with Google
复制标题
使用网络分析重新排序搜索结果与谷歌的案例研究：与谷歌的案例研究
DOI：
--
复制
发表时间：
2002
期刊：
影响因子：
--
通讯作者：
M. Chignell
中科院分区：
文献类型：
--
作者：
Behnak Yaltaghian;M. Chignell
在本文中，我们审查的方法结构化搜索信息在万维网上。我们提出了新的方法的基础上共引和网络分析。我们描述了一组21措施，这些方法的基础上，并检查这些措施的因素结构。然后，我们报告了我们最近在多伦多大学进行的一项研究。人工评判员对谷歌搜索引擎返回的七个查询中的每一个的相关性进行了评分。我们比较了Google选择的前20个搜索结果与21个网络分析指标中的每一个选择的前20个结果的平均判断相关性。除了一个网络分析测量（“inlink”）之外，所有网络分析测量都显示出在他们的前20个选择中显著（p<0.05）更好的（与谷歌相比）平均判断相关性。逐步回归分析，然后用三个网络分析措施作为预测因子，这占了大约17%的相关性判断的方差确定一个线性模型。虽然这些结果需要通过对广泛的查询和主题进行更详细的分析来扩展，但它们表明搜索输出邻接矩阵的网络分析（其中邻接/邻近基于网络范围的共同引用）可能会显着提高搜索引擎排名。
In this paper we review methods of structured search for information on the World Wide Web. We propose new methods based on co-citation and network analysis. We describe a set of 21 measures based on these methods and examine the factor structure of those measures. We then report on a recent study that we have conducted at the University of Toronto. Human judges rated the relevance of a selection of Web pages returned by the Google search engine for each of seven queries. We compared the average judged relevance of the top 20 search results selected by Google vs. the top 20 results as selected by each of the 21 network analysis measures. All but one of the network analysis measures ("inlink") showed significantly (p<.05) better (as compared to Google) average judged relevance amongst their top 20 selections. Stepwise regression analysis was then used to identify a linear model with three network analysis measures as predictors, which accounted for roughly 17% of the variance in relevance judgments. While these results need to be extended with more detailed analysis of a wide range of queries and topics, they suggest that network analysis of search output adjacency matrices (where adjacency/proximity is based on web-wide co-citations) may significantly improve search engine rankings.