Link analysis ranking

摘要: The explosive growth and the widespread accessibility of Web has led to surge research activity in area information retrieval on World Wide Web. Ranking always been an important component any system. In case search its importance becomes critical. Due size Web, it is imperative have ranking functions that capture user needs. To this end offers a rich context which expressed through hyperlinks. thesis we investigate, theoretically experimentally, application Link Analysis Building upon framework hubs authorities [57], propose new families algorithms. Some algorithms define no longer enjoy linearity property previous As result, harder analyze them, or even prove they actually converge. However, for special consider, are able will converge, provide complete characterization combinatorial properties stationary authority weights produces. The plethora generates necessity formal way evaluate their compare behavior. We introduce theoretical study algorithms, specific within framework. Using these axiomatic INDEGREE algorithm ranks pages according number in-coming links. We conclude with extensive experimental evaluation Ranking. test over multiple queries, use feedback determine quality. Our experiments reveal some limitations Specifically, appears most nodes structures graph favor, do not correspond relevant collection. These observations offer insight into mechanics believe lead improved design, better input graphs