Chapter 1 Adaptive Information Filtering

A financial analyst wants to be alerted of any information that may affect the price of the stock he is tracking; an agent working in the Homeland Security Department wants to be alerted of any information related to potential terror attacks; a customer call center representative wants to answer customer calls about problems that he can handle; and a student wants to be alerted of fellowship or financial aid opportunities appropriate for her/his circumstances. In these examples, the user preferences are comparatively stable and represent a long term information need, the information source is dynamic, information arrives sequentially over time, and the information needs to be delivered to the user as soon as possible. Traditional ad hoc search engines, which are designed to help the users to pull out information from a comparatively static information source, are inadequate to fulfill the requirements of these tasks. Instead, a filtering system can better serve the user. A filtering system is an autonomous agent that delivers good information to the user in a dynamic environment. As opposed to forming a ranked list, it estimates whether a piece of information matches the user needs as soon as the information arrives and pushes the information to the user if the answer is " yes " , so a user can be alerted of any important information on time. A typical information filtering system is shown in Figure 1.1. In this figure , a piece of information is a document. A user's information needs are