Text Mining in Real Time

Tuesday, April 18, 2000 - 1:30pm - 1:55pm
Keller 3-180
Henry Lieberman (Massachusetts Institute of Technology)
There is a tendency to try to apply the old Information Retrieval paradigm to the Web. This views the Web as a [relatively] static database, to which the user issues a query that expresses what he or she wants and the goal of a search engine or agent is to return the best document in answer to the query. But the Web is different. Users dynamically browse through the Web, they have difficulty expressing precisely what they want, their needs, interests, and context change quickly, and the Web itself is a dynamic entity. A different approach is to view Web exploration as a cooperative activity between [one or more] human users and [one or more] software agents, acting together in real time. The goal is really to make the best use of the user's time. I will describe some software agents that try to learn the user's interests automatically from observation, proactively anticipate user needs, and act as advance scouts during Web exploration. They provide recommendations continuously in real time. This leads to a style of interaction that is neither purely browsing nor purely searching, but combines aspects of the two.