Help Test an "Adaptive" Search Engine

An experimental science and medicine search engine needs your help with its ambitious goal of automatically improving search results by observing user behavior.

Phibot is a research project of the University of Mainz and the German Institute of Artificial Intelligence. The system uses some nifty technology that literally helps the engine automatically learn the difference between "good" and "bad" results, over time.

"We try to automatically adjust the search engines algorithms to the specific user information needs, to provide better search results," says Phibot representative Thorsten Henninger. "We test out different methods to achieve this, [such as” query refinement and personalized search algorithms."

The search engine crawls both science and medical web pages and news sources, offering an interesting mix of information. Unlike nearly every other search engine, it also allows you to choose the query processing algorithm used to determine results.

You can select a "conventional" algorithm (which the site says is like Google's), or three different "vectorspace" models, typically used by more traditional search systems.

Phibot has about 70 million pages in its medical science index, and about 200 million general science pages, making it a very useful science-specific search engine. But the researchers behind the project aim to make it even better over time.

"We still need more search requests to get a deeper understanding of the users search behaviour," says Henniger. "Phibot works with SessionIDs (like Amazon). Therefore we can identify a whole search session of a user by those SessionIDs."

You can also explicitly "vote" on the quality of the results, using the "ClickThroughEvaluation" feature. Each result is accompanied by a "good" and "bad" radio button that you use to submit feedback on the quality of the results. A "scan" feature opens each document as a simple text file in its own window, making it easy to quickly view each result before voting.

What's in it for you, other than the opportunity to help improve Phibot? Data.

Phibot updates all of the data it collects once a week, and makes it available for downloading at no charge. All logfile information from the search engine is available for research purposes in an XML-style file format.

This information can be a gold mine for studying how people search. You can use this information to improve the overall search usability of your own site, even if its primary subject is not science or medicine.

Phibot's ultimate goal is ambitious: to create a search system that automatically "improves" over time by observing the behavior of its users. Beyond being a useful science and medicine search engine, Phibot's efforts toward improving search are worthy of your support.


Phibot ClickThroughEvaluation
Use this link to have search results appear with radio buttons you can use to evaluate the quality of results.

Search Headlines

NOTE: Article links often change. In case of a bad link, use the publication's search facility, which most have, and search for the headline.

Domain name news
Hanna Barbara wins Scooby Doo domain name appeal...
Demys Sep 11 2002 12:39PM GMT
Online portals news
Yahoo set to take speed challenge...
CNET Sep 11 2002 11:14AM GMT
Yahoo, ISPs dig in against labels...
ZDNet Sep 11 2002 8:42AM GMT
Online search engines news
AltaVista and Google to fight Chinese censorship...
The Register Sep 11 2002 7:30AM GMT
Online portals news
With Ad Revenue Down, Yahoo Tries to Win Agencies' Favor...
New York Times Sep 11 2002 6:18AM GMT
Top internet stories
Study: 20% of U.S. Feels Web Is 'Essential'...
Internet News Sep 11 2002 5:37AM GMT
Online content news
New York Times Digital Tests Paid Content Sales, Here's What They Learned... Sep 11 2002 1:37AM GMT
Online search engines news
FavouriteWebsiteAwards (FWA) Launches To Spotlight Best Web Design...
URLwire Sep 10 2002 10:36PM GMT
Online marketing news
Web sites to limit ads on 9/11...
CNET Sep 10 2002 8:16PM GMT
Web developer news
Robots, Agents and Spiders - Identifying Search Engine Crawlers...
Stickysauce Sep 10 2002 1:58PM GMT
Online marketing news
Web pop-up merchants named and shamed... Sep 10 2002 12:36PM GMT
Online portals news
AOL woes could bruise Yahoo...
ZDNet Sep 10 2002 12:26PM GMT
Domain name news
Domain firm outlines '' rules...
ZDNet Sep 10 2002 12:26PM GMT
powered by

About the author

Chris Sherman is a frequent contributor to several information industry journals. He's written several books, including The McGraw-Hill CD ROM Handbook and The Invisible Web: Uncovering Information Sources Search Engines Can't See, co-authored with Gary Price. Chris has written about search and search engines since 1994, when he developed online searching tutorials for several clients. From 1998 to 2001, he was's Web Search Guide.