InventorEd, Inc.

How Good Are Search Engines?

Click here to break out of Frame

Blue bar graphic divides sectins of the page

In the April 3rd SCIENCE, Steve Lawrence and C. Lee Giles report on their investigation of coverage by web browsers.  They estimate the web has 320 million indexable pages. Coverage the browsers they examined were:

                                Coverage        95% Confidence
                                                                Interval

HotBot                  57.5%              +/- 1.3
AltaVista              46.5                  +/- 1.3
Northern Light     32.9                  +/- 1.1
Excite                   23.9                  +/- 0.86
Inforseek              16.5                  +/- 1.0 
Lycos                     4.4                   +/- 0.42

The authors, quite reasonably conclude, that if you want to have the best results from searching, it pays to use multiple browsers or a softbot web search intelligent agent.

Our experience is that Alta Vista is pretty good for searching for scientific and technical material as well as product literature. But we too find that just relying on Alta Vista causes you to miss important sites. So we usually begin by conducting a Metacrawler search (
http://www.metacrawler.com/). We use that search to identify which search engines seem to be producing the best results and then use those engines for more complicated queries which cannot be supported by metacrawler and other search engines which browse on the results of other engines.

Another approach, implemented internally at Foresight, is to store useful sites by profiles. That is, each site identified in the firm in connection with commercialization work is stored in a data base together with motivations for searching and information on the technology, applications, markets, and industries involved (the profile). A case based reasoner allows
us to rapidly retrieve sites found useful on prior projects.  If you figure it often takes between five and 15 minutes to find a highly useful site, the ability to cut that search time to seconds has clear utility.  You can make a simple variant of this tool yourself by organizing your "favorites" in your browser  into appropriate folders. Folders can be attached as e-mails
for sending to colleagues.

Good Hunting!
Phyl

Techlist is a moderated listserver for discussions and questions concerning the commercialization of technology and technology transfer.  To join or leave techlist send an e-mail to list@seeport.com with the appropriate command in the first line:

JOIN techlist (the command SUBSCRIBE may also be used).
LEAVE techlist (the command UNSUBSCRIBE may also be used).

All messages to Techlist, except spam or flames, are posted for all members.
Opinions expressed are those of the authors. Administrative e-mail should
be sent to the moderator: 
phyl@seeport.com

***********************************
If opportunity doen't knock, build a door.
Milton Berle

Phyl Speser, J.D., Ph.D. , List Moderator
Foresight Science & Technology,
1200 W. Sims Way, Suite 201, Port Townsend, WA  98368
Voice: 360.385.9560, ext. 11, Fax: 385.998
phyl@seeport.com
http://www.seeport.com

Blue bar graphic divides sectins of the page
Have you been Scammed? Find out what to do here.
PLEASE: Notify Us Of Any Errors Of FACT:
ErrorOfFact@InventorEd.org

Blue bar graphic divides sectins of the page

Site Index - Contact Us - © InventorEd, Inc., Page last revised 10-14-2000

Graph of the electromagnetic spectrum