New search tool for CUinfo web information is agile, adaptable, accurate

David Baer, a senior in computer science, investigated the use of CUInfo's new search tool, Ultraseek. Nicola Kountoupes/University Photography

By Beth Goelzer Lyons

A powerful new search engine is making it easier -- much easier -- to find information on 200,000 of Cornell's web pages. Behind CUinfo's search feature is Ultraseek, a cousin of the popular Internet search tool Infoseek (also known as the GO Network).

Before Ultraseek, searches on CUinfo were handled by an outdated, sluggish tool called Harvest. Last year, David Baer, a Cornell senior in computer science, began researching alternatives. Baer has worked on CUinfo's search tools, located at http://www.cornell.edu/CUsearch.html, since his freshman year.

"Harvest was designed to handle a much smaller number of web pages than we have now. It was very slow and often broken," explained Baer. "So we investigated other options and looked at what other institutions similar in size were using. Ultraseek came out on top for many reasons."

One reason is its speed. Enter a search request and the results will be back in less than a second, on average. Harvest took about 10 seconds -- long enough to make people think it wasn't working and give up.

Ultraseek's flexibility is another plus. Complex search requests can be made by completing a short form or by composing a search string in the traditional way. Ultraseek initially displays the results by how closely they match what the user likely wanted. One button click redisplays them from newest to oldest. To further tailor the search, a user can select one result and tell Ultraseek to find other pages like it.

By contrast, search results in Harvest were quite limited -- only the first 50 matches would be listed and in no particular order. Harvest also failed to find pages that it should have found because its index was often out-of-date.

All search tools rely on an index of information gathered from web pages. When a user requests a search, the index is what is actually checked, not the web pages themselves. In building its index, Harvest visited each page one by one, a process that took up to 14 days. Ultraseek cuts its index building time to just three days by visiting multiple pages simultaneously.

The Ultraseek index currently comprises the CUinfo web server and other Cornell-affiliated servers that have registered with CUinfo. For information about registering, e-mail CUinfo-admin@cornell.edu.

Easy search tips:

• Click on the "Advanced" link underneath the search box if you need to do a complicated search.
• Put quotation marks around words that must go together, such as "Cornell Dairy." Without quotes, Ultraseek would list all pages that match "Cornell" and/or "Dairy."
• Use several words to describe exactly what you want. Your search results will be better.
• Use capital letters if you are searching for a proper name that could also be a common word.
• Put a plus (+) sign at the beginning of words that Ultraseek must find. Put a minus (-) sign in front of words that Ultraseek should exclude.