Marc Najork

"Comparing the Effectiveness of Different Scoring Functions for Web Search"

Wednesday, February 14, 11 AM

Maginnes - Room 101 

In this talk, I compare the effectiveness of several popular scoring functions used for ranking web search results: PageRank, HITS hub and authority scores, web page in- and out-degree, and BM25F (a scoring function based on textual feature of a web page and anchor text referring to it). I use three popular performance measures to compare these scoring functions: the mean reciprocal rank, the mean average precision, and the normalized discounted cumulative gain.  The evaluation is based on two data sets: a large web crawl of 463 million pages, and a set 28043 queries drawn at random from the Live Search query log, with results labeled by human judges as to their relevance.

Marc Najork is a Principal Researcher at Microsoft Research Silicon Valley.  Marc has a long-standing interest in Web research. He has worked on link-based ranking algorithms, heuristics for detecting "spam" web pages, high-performance web crawling, studying various structural properties of the Web, and constructing frameworks that make it easy to build distributed web-based applications.  Marc served as program co-chair of the 13th International World Wide Web Conference (WWW 2004), and is on the editorial board of the ACM Transactions on the Web (TWEB).  He joined Microsoft in 2001, after having spent eight years at DEC's (later Compaq's) Systems Research Center. Marc received a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign.  His home page is at http://research.microsoft.com/~najork/