WebWatcher: A Learning Apprentice for the World Wide Web

We describe an information seeking assistant for the world wide web. This agent, called WebWatcher, interactively helps users locate desired information by employing learned knowledge about which hyperlinks are likely to lead to the target information. Our primary focus to date has been on two issues: (1) organizing WebWatcher to provide interactive advice to Mosaic users while logging their successful and unsuccessful searches as training data, and (2) incorporating machine learning methods to automatically acquire knowledge for selecting an appropriate hyperlink given the current web page viewed by the user and the user’s information goal. We describe the initial design of WebWatcher, and the results of our preliminary learning experiments.