- OSU
News Research Archive
(an archive of past stories)- Coverage of OSU Research
- Reports on national news stories
- OSU Cancer Report
- (cancer research and treatment)
- Frontiers
- (a magazine about cancer research and treatment)
- Synergy
- (a magazine produced by the College of Biological Sciences)
- Reporting on Cancer
- (a reporter's guide to the disease)
- Science Communications Staff
Who we are and what we do.
(Last updated 12/3/99) - Coverage of OSU Research
11/30/99
STUDY: INFORMATION ON THE WEB IS LIKELY CORRECT, BUT HARD TO FIND
COLUMBUS, Ohio -- People who search for answers to questions on the World Wide Web are more likely to find the right answer than a wrong one, according to an Ohio State University scholar. The hitch: most likely, they won’t find an answer at all.
A study found that the Internet search engine AltaVista (http://www.altavista.com) uncovered Web pages with the correct answer to library reference questions 27 percent of the time, and wrong answers 9 percent of the time. However, 64 percent of the time the Web pages it listed either contained no answer or were out of service.
The study suggests that people shouldn’t use the Web as their only source of information, said Tschera Harkness Connell, serials coordinator at the Ohio State University Libraries. They can, however, supplement other resources with information from the Web with some success.
“The best advice I can give is to evaluate the credibility
of sources, and double-check the facts you find on a Web site
with other information sources whenever possible,” Connell
said. If Web sites list sources’ credentials and offer
references to where they found the information, she said, the
site is more likely to be trustworthy.
“The great thing about the Web is that it enables anyone with access to it to publish their thoughts,” Connell said. “That’s also the worst thing about it. Not every person is publishing accurate information.”
Connell completed this study while on the faculty of Kent State University, with Jennifer Tipple, then her student. Tipple is now a reference librarian at the Dayton and Montgomery County Public Library in Ohio, and Connell is continuing this work at Ohio State. The study appeared in a recent issue of Reference & User Services Quarterly.
For the study, Connell and Tipple evaluated AltaVista’s performance at finding sites to answer 60 questions posed by library patrons. The questions came from patrons of one suburban library during a one-week period. They chose questions with relatively simple answers that they could find in standard reference books such as encyclopedias or almanacs. Tipple chose the AltaVista search engine because it is highly rated and she was familiar with it.
Some questions patrons asked:
- Who was Annie Oakley, and for what was she known?
- What is the population of Columbus, Ohio?
- Why do leaves change color in autumn?
- How do you spell Deion Sanders’ first name?
When Tipple searched for keywords related to these questions on AltaVista, 64 percent of the Web pages it returned either didn’t contain the answer or were no longer working. About 9 percent contained an incorrect answer, and 27 percent contained the correct answer.
Tipple re-worded and re-entered searches when AltaVista returned no relevant Web pages within the first two screens of results. For 60 questions she had to perform a total of 106 searches, an average of 1.77 searches per question.
Connell said people’s choice of keyword is very important to finding the right answer. She offered as an example the question “How do you spell Deion Sanders’ first name?” Typing in a guess such as “Deon” returns more than a hundred sites where the Web page author, like the person doing the search, spelled Sanders’ name incorrectly. (Sanders is a player with the Dallas Cowboys football team.)
The situation is even more complicated, Connell said, because Web page owners can pay some search engines to have their page featured at the top of a list of results for a particular keyword. When this happens, their page turns up even when it isn’t particularly relevant.
Some 21 percent of Web pages returned in this study were duplicates. AltaVista doesn’t employ staff to remove such errors, a fact that led Connell to say the number of duplicates in its database will probably increase over time.
Connell contrasted AltaVista with Yahoo!, another popular search engine. Yahoo! does employ a staff to review each site, but that takes time, so Web pages aren’t available to the public as soon as they are registered, as they are with AltaVista.
Ultimately, the best way to use the Web is in conjunction with other sources of information, Connell said.
“An encyclopedia will readily answer the question, ‘Who is Annie Oakley?’” she said. “But on the Web you may find a person whose hobby is knowing everything there is to know about Annie Oakley. Their site may be able to provide you with very interesting and colorful information that is credible, but isn’t in the encyclopedia.”
Currently, Connell is investigating to what extent following up “no answer” results with a second search engine increases the chance of finding correct information. She is also studying whether using browsers that rely at least partially on human indexes enhances success.
- Contact: Tschera Harkness Connell, (614) 292-6314;
Connell.17@osu.edu- Written by Pam Frost, (614) 292-9475; Frost.18@osu.edu