Improving ranking -Reply

Martin Courtois courtois at gwis2.circ.gwu.edu
Mon Jun 22 09:19:43 EDT 1998


I've been working with a faculty member in the Computer Science dept. at
University of Tennessee, Knoxville to test the ranking capabilities of
major WWW search engines.  We just finished analyzing 6000 pages retrieved
by searches and we're surprised at the results.  Even for a very simple
test, such as "Are any pages that do not contain all search terms ranked
higher than pages that do contain all search terms?," search engines
showed error rates of 8 to 30 percent for the first 100 items retrieved. 
Tests for proximity of terms and location of terms (meta tags, title,
headings, etc.) were even worse, with error rates of up to 80%. 

With the dynamic nature of the Web, secrecy of ranking algorithms, and
spamming/spoofing techniques, it's difficult to achieve consistent results
from search engines.  I think Danny Sullivan (www.searchenginewatch.com) 
offers the best advice for techniques to achieve the best ranking possible
without turning this into a full-time job or costing way too much.

Several people have commented on search engines' use of meta tags.  Keep
in mind that Excite ignores all terms in meta tags.  Lycos indexes these
terms, but they are treated as any other text, i.e., they do do not
receive a relevancy boost. 

Marty    

***********************************************************
Martin Courtois
Electrical Engineering/Computer Science Reference Librarian
Gelman Library
George Washington University
2130 H St., N.W.
Washington DC 20052
Phone: (202) 994-0684
FAX: (202) 994-1340
E-mail: courtois at gwu.edu



More information about the Web4lib mailing list