Professors moonlight as Web entrepreneurs

Peter Kauffner

Professors don’t spend all their time teaching and researching — some are starting technology corporations.
Net Perceptions, founded last July by computer science professors and entrepreneurs John Riedl and Joe Konstan, has developed a product to guide users through the World Wide Web based on their personal preferences.
The product, called the GroupLens Recommendation Engine, uses a technique developed by Riedl called collaborative filtering. Software users fill out a questionnaire, and the program makes recommendations according to what previous users reported with similar preferences.
The idea evolved from a program completed in 1994 to filter Usenet, the Internet’s main discussion board.
“We wrote this program that lets people rate the articles they read and help other people find the articles they like,” said Riedl, now the company’s chief technical officer.
The program is available for public use as part of a University research project.
And it’s not the only entry into the collaborative filtering market. A team from the Massachusetts Institute of Technology founded a competing company, Firefly Network, Inc., in 1995. But both companies are still feeling their way through a fledgling industry.
“I don’t know how you make money off of Usenet,” Riedl said. “We’re doing (this) because Usenet is just screaming for someone to pick out the good stuff from all the junk.”
But when booksellers, travel agencies and others begin to use the Web as a marketing tool, Riedl and his colleagues began to think that the idea could be adapted for commercial use.
After venture capitalist Ann Winblad agreed to provide a first round of financing last June, the new company was launched. The company also received $4 million from three venture capital companies in December. But the University holds the patents on the professors’ work.
“We wanted to do it all by the book with the University licensing the technology,” Konstan said.
One of Net Perceptions’ first customers was, a company that sells books through the Web.
“They’re actually not live on their site with it yet,” Riedl said. “The idea is that you will be able to say which books you like and which you don’t like and then its says, ‘Well, you should consider these books because people who are similar to you liked them.'”
Customers will then rate these recommendations, allowing the program to make recommendations that are specifically tailored to the customer’s interests.
“It’s like joining a book club,” Riedl said.
The program also tests for behavior that might indicate a preference.
“What we tested in (Usenet) is that the amount of time a person spends with an article is a pretty good indicator of whether they find it interesting,” Konstan said. Net Perceptions has a patent pending, independent of the University, on this technology, called “implicit ratings.”
Another application Riedl hopes the program will have potential with is a new area called “push vendors.”
Such vendors download files to a subscriber’s computer at night or at times when the computer would not be in use.