WebWatcher: A Learning Apprentice for the World Wide Web

Abstract

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. 6 Machine Learning Information Services Tlds ~ti~ being rnaintu~cd bF the ML Groul~ at the AustrianResea~.h Institute [or Arti0cial Int cll~ence (OPAl), Vienna, Austria It is far from complete and is being opdatcd on anin'cgular basis. Pleas e direct comment/ su~estions/...to Gelll~d Widmer {gerh~dCc~,cd,~nil'ie,¢~:.) To ~' out our experlnncnt~l WcbWatcher s catch as slstant, click lj_ _e~_ GeJmral ML l,tfot'trtatio,t Sotu'ces

How to Build Modeling Agents to Support Web Searchers

“…A specific user model can be based on explicit conversations with an agent (Etzioni and Weld, 1994), or as a mixture of explicit interactions with an agent and interactions with the web itself (Armstrong et al, 1995;Thomas and Fischer, 1996). By contrast, our agents were designed to operate in a style we believe is better suited to assisting users in their normal use of the web.…”

Personalized Paper Recommendation Based on User Historical Behavior

“…The earliest recommendation system was based on content-based filtering including Web Watcher [10], LIRA [11], Leticia [12] and et al All of them recommended resources by evaluating the similarity between resource content and user interest.…”

WebSA: Database Support for Efficient Web Site Navigation

“…Software tools sometimes called "agents" are being deployed for speeding up the information retrieval process on the web in general [1,3,4,6,9,10,11,16,17] and within individual web sites in particular [7,12,13].…”

“…In some systems, the user is also asked to give feedback on the retrieved documents [3,4] in some cases for the purpose of training a learning algorithm [1].…”