In this project Filippo Menczer and his researchers study the relationships between different types of Web topology based for example on hyperlinks, words, and page meaning, and how they affect the performance of ranking and crawling algorithms, such as InfoSpiders. This research extends prior work in which they characterized a necessary condition for effective autonomous browsing of any distributed hypertext database such as the Web in terms of a relevance autocorrelation measure. More recently they have used a brute force approach to map the relationships between lexical, linkage, and semantic similarity across billions of Web page pairs. This research is being applied to build models that may help understand how the scale-free distribution of Web links has emerged and how it can be exploited for designing more effective Web crawlers and search engines. This has been added to the articles section of Deep Web Research Subject Tracer™ Information Blog.
posted by Marcus Zillman |
4:15 AM
