Improved Relevance Ordering for Web SearchJanuary 1, 2003
Overture Services, Inc. Mathematics, 2002-03
Liaison(s): Dan Fain
Advisor(s): Leslie Ward
Students(s): Erin Bodine (PM), David Gleich, Cathy Kurata, Jordon Kwan
We implement and compare the performance of three algorithms for reranking the top 50 webpages returned by a search engine for 40 individual queries. We then compare these re-rankings to human rankings. Two algorithms break webpages into topically distinct areas (Micro Information Units). The third computes a proximity score for each webpage, measuring how close the words in the query appear. Query terms appearing closer together may indicate the webpage is more relevant to that query.