inconsistencies web search performance

Dorothy Day day at indiana.edu
Fri Nov 7 22:00:57 EST 1997


On Fri, 7 Nov 1997, N. TOMAIUOLO, INSTRUCTION LIBRARIAN, CCSU wrote:

> 
> The point is that as a professional searcher who has seen the accuracy of 
> powerful engines including Dialog, BRS, STN and reliable CDROM products, 
> I'm wondering when we are going to be offered this type of accuracy
> with the web.
> "The Web made simple" book, and "Search Engine Secrets ofthe Pros" (PC World 
> article) nothwithstanding -- these engines including AltaVista, Lycos,
> InfoSeek, et al, do not perform as they should on a consistent basis.
> Is this vexing anyone besides me?
> 

Most of the postings along this line seem to be comparing
controlled-vocabulary searching on commercial databases with the
full-text searching common to most web search engines, a rather dubious
approach. It would make more sense to compare a free-text search of
full-text items in a (say) Dialog database with a similar search on the
web.  Many of the same problems arise--and many of the same attempts are
made at solving them: relevance ranking schemes, use of WITHIN and NEAR
operators, ways to indicate phrase terms, etc. Nowhere has full-text
searching allowed the same control as one can get searching controlled
vocabulary in specified fields, but searchers still use the former to
extend the richness of the search beyond the limited range of indexed
terms. 

Yes it would be nice to have more controlled vocabulary tools available
for web-indexed resources, but someone will have to pay to have that
done, and it won't come cheap. With the volume of material added daily,
I wonder what kind of automated indexing could keep up and generate the
kind of indexing we get in much smaller databases of more even quality
and limited range of coverage. Nothing available to date, I don't think.

*****
Dorothy Day			
School of Library and Information Science
Indiana University
day at indiana.edu	
*****
	"He also surfs who only sits and waits."





More information about the Web4lib mailing list