The problem

The number of networked users has increased rapidly with the widespread proliferation of computers and networks. If the Internet is any indication, the number of people who have started using online services has increased dramatically in recent years, . The number of Internet hosts, i.e. machines which have direct connectivity, is over a million [26] and is growing exponentially. There are many more which connect to the Internet indirectly through intermediary services such as America Online and Compuserve. This explosive growth has fed the growth in the amount of information resources available over the networks. As more information becomes available, it becomes increasingly difficult to search for information. However, as the number of users increases, newer users are likely to be less network savvy. Getting information should become easier, not harder, if newer users are to be able to meet their information needs. It is, therefore, critically important to build tools that help users serve their information needs better.

Information Filtering deals with the delivery of information that is relevant to the user in a timely manner. An information filtering system assists users by filtering the data stream and delivering the relevant information to the user. Information preferences vary greatly across users, therefore, filtering systems must be highly personalized to serve the individual interests of the user. A personalized filtering system must satisfy three requirements:

• Specialization: A personalized filtering system must serve the specific interests of the user. The system selects the articles deemed to be interesting to the user and eliminates the rest. However, a filtering system might not be able to perfectly differentiate the articles that are actually relevant to the user from the ones that are not. The proportion of irrelevant articles delivered to the user should be as low as possible. The proportion of relevant articles eliminated should also be low. Since filtering involves repeated interactions with the user, the system should be able to identify patterns in the users behavior. The filtering system must infer the habits of the user and specialize to them, i.e. recommend as many relevant articles and as few irrelevant articles as possible.

• Adaptation: Since filtering typically involves interaction over long periods of time, user interests cannot be assumed to stay constant. When they change, the system must first be able to notice that the user interests have changed. Secondly, the system must adapt its behavior in response to the change. Anticipating and adapting to user needs helps make the system more user friendly. This is essential if more and more people are to use it.

• Exploration: A filtering system should also be capable of exploring newer information domains to find something of potential interest to the user. There are two motivations for exploration. One is that exploration helps match a presently unknown but real user interest. The other motivation is that it helps improve the adaptation process. This is the case because newer kinds of information need to be explored to serve the changing user interests.

On the one hand, users need to be able to control the huge amounts of information inflow and fulfill their information needs. On the other hand, users need to be relieved of repetitive computer tasks by providing higher level interface abstractions. A solution for both problems is to design an information filtering system that learns. The goal of an information filtering system is not to fully perform all information filtering tasks, but to automate those that are more repetitive and predictable. Filtering systems can perform some of the filtering tasks and help the user manage the flux of information. Learning filtering systems can adapt to the tastes and preferences of the user and automate tasks for them. The proposed information filtering agents will be analyzed in detail in the following chapters.