Sniffing out the best of the Web
In an ideal world, everyone on the Internet would have a personal "filtering agent" to sort through the thousands of documents that come on-line daily and pinpoint those of particular interest. But how would this work? One possible solution is to build a personalized agent that automates the transfer of information and makes recommendations to users with similar interests.
Enter Webhound, a World Wide Web document recommendation service being developed by Yezdi Lashkari, a Media Laboratory graduate student working with Professor Pattie Maes in the autonomous agents group. By using a sophisticated evaluation system, Webhound automates the process of "word-of-mouth" recommendations for Web users by creating a personalized profile for each user based on significant features of documents on which they provide opinions. It then matches their profiles with profiles and recommendations from a universe of other users. The more document evaluations each user gives Webhound, the higher the quality of subsequent personal recommendations.
The system relies on its users to submit opinions on new documents they discover while surfing the Web, and the database of documents is expanded entirely by its user community. Webhound averages 10 to 12 new users daily, but Lashkari stresses that it's "still very much a work in progress. We haven't solved all the hard problems yet."
If you are interested in learning more about Webhound, you can find it at
Toward deregulating electric utilities
To increase efficiency and decrease costs, congressional actions and economic market forces are together rapidly moving the electric power generation sector from regulation to competition.
But running a regional electric power network involves making short-term decisions and moment-to-moment adjustments in generation to ensure that electricity keeps flowing to customers at an acceptable frequency and voltage. If the regulated regional utilities are replaced by dispersed, competitive generating companies, who will run the interconnected power network?
MIT researchers led by Dr. Marija Ilic of the Department of Electrical Engineering and Computer Science and the Energy Laboratory have outlined an approach that uses existing operating systems and data both to control operation of the network and to establish prices that reflect not just generating costs but also the changing costs incurred in maintaining the network. The result of their approach should be an industry that provides electricity efficiently, reliably, and at prices that encourage consumers to use electricity in a cost-effective manner. The work has been sponsored in part by New England Electric System and the DOE. (Source: e-lab, a publication of the MIT Energy Laboratory)
A version of this article appeared in MIT Tech Talk on May 17, 1995.