🚀 go-pugleaf

>Evaluating individual resources is all very well, but for many queries the
>first problem is reducing the number of candidates. Faced with 10000
>replies one should of course elaborate the query to try to reduce the
>number,  but even when the response has fallen to say 256 there is still a
>problem. If one would be content with approximately one answer, it is very
>likely (but, pax to some contributors, not certain) that adding AND NOT
>url:com to the query will cut the number to say 30 or 40. One may have
>thrown out a jewel, but ...
                        <-SNIP->

>It would be interesting to know what other weeding techniques members use.

>-----
>Nick Capon          (08) 8278 3584
>retired from CS Dept, Adelaide U., South Australia
> mailto:ncapon@adelaide.on.net

Actually, since most of the search engines I train with use
relevance ranking, I don't worry too much about the colossal number of hits
that a given search may return.  In fact, I tell them to plug in as many
relevant search terms as they can think of (or between 5 and 10).
The search engine will automatically place the pages with the most appearances
of those words at the top of the list.  Sure there may be 965,124 hits, but
you are only going to care about the first dozen or so.

It's not foolproof, but it is a way to get the trainees quickly searching, and
they learn best by doing. They tend be more interested in results than in the
search.  Myabe this is good, maybe its bad, but so far it is working pretty
well.

Steve Bachman   (4CTY_STEVE@4cty.org)
Electronic Services Librarian
Four County Library System
304 Clubhouse Road
Vestal, NY  13850
http://lib.4cty.org