Over the last decade, there has been vast amount of research in automatic mediation of access to networked information. As a result of this research, a number of methods for information filtering have been presented. Information Filtering (IF) is an area of research that offers tools for discriminating between relevant and irrelevant information by providing personalized assistance for continuous retrieval of information. IF is needed in situations of information overflow in general, and on the Internet, in particular. This area combines tools from the field of artificial intelligence (AI), such as intelligent agents or software robots (“softbots”), guided by user profiles, with information retrieval (IR) methods, geared to the representing, indexing and retrieving of content. Agent technology provides a framework for automated information gathering over the Internet. Indeed, quite a few applications have been developed for this purpose, for example, passive filtering of incoming messages, like email and Usenet data. Active information seeking, like interesting web-site detection and browsing assistance, presents another application. The heart of such an agent is the “user profile” - a representation of user needs, which is constantly updated according to user feedback. The performance of IF systems (namely, their ability to retrieve or filter relevant information) depend heavily on the accuracy (how well it represents user interests) of the user profile. Problems related to user profiles includes, how to generate an initial profile for a new user, and how to update an existing profile over time.
We plan to compare and evaluate several methods for user profile generation and update including of “content based” and “rule-based”. The methods that we will examine include: content based user-defined profile, system (automatic) defined profile based on automatic indexing, system (automatic) defined profile based on artificial neural-network, rule-based personal profile, and rule-based stereotypic profile and their combination. We plan to develop an IF agent that will employ the above methods. The agent will enable to generate an initial user-profile and to update it later on, based on user’s relevance feedback. The agent will utilize the most effective method for generation and update of user profile as discovered in this research.