home · mobile · calendar · colloquia · 1999-2000 · 

Colloquium - Shavlik

Creating an Instructible and Self-Adaptive Web Browser
Computer Sciences Department, University of Wisconsin - Madison

Web browsers have become major user interfaces, providing access to a nearly overwhelming amount of information. We are developing machine-learning systems that strive to increase the intelligence and value of web browsers.

In one project we are working on widening the "communication pipeline" between human teachers and their learning systems, accommodating a richer human-machine dialog than the traditional input-output pairs. We are constructing an instructable software agent that wanders the web and finds pages of interest. The Wisconsin Adaptive Web Assistant (WAWA) accepts "advice," with which human users specify their interests and provide hints on how to find relevant information. WAWA compiles these instructions into neural networks, which means that the system's behavior can be further modified via training examples. Users can create these training examples by rating pages retrieved by WAWA, but more importantly the system uses techniques from reinforcement learning to internally create its own examples (users can also later provide additional advice).

In a second project, we are investigating which measurements routinely collected by current operating systems and web browsers can serve as reliable indicators of a user's interest in a web page. Such measurements can be used to automatically create labeled training examples, circumventing the need for users to tediously rate pages according to their interestingness.

Jude Shavlik is an ICS Distinguished Speaker.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)