Representing Users in a Travel Support System Maciej Gawinecki, Zygmunt Vetulani Faculty of Mathematics and Computer Science Adam Mickiewicz University, Pozna , Poland vetulani@amu.edu.pl Minor Gordon Department of Computer Science Technical University of Berlin, Berlin, Germany minorg@cs.okstate.edu Marcin Paprzycki Department of Computer Science OSU, Tulsa, OK, USA and Computer Science, SWPS, Warsaw, Poland marcin@cs.okstate.edu Abstract We consider the construction and management of user profiles for an agent-based travel support system, with the goal of providing personalized content for individual users of the system. Profiles consist of statements about the “world of travel.” Conditional probabilities for each statement model the strengths of user preferences. These probabilities are derived from implicit and explicit observations of user behavior. 1. Introduction The ability of sites such as Travelocity and Expedia to filter the mass of travel-related information on the Internet is key to their popularity. Content filtering and personalization is the antidote to information overload, and as such it plays an increasingly important role in our interactions with large-scale databases such as airline reservation systems and yellow page listings. Software agents have long been touted as a facilitating technology in this respect. The natural mapping between travel agents and software agents has inspired a number of attempts to design and implement an agent-based travel support system. Although most of these designs were purely conceptual, a few have proceeded to the implementation phase. We have developed an agent-based system for providing personalized travel-related content to Web users [2, 9, 13, 14, 24, 25]. In a recent paper [15] we have summarized current design of the system, which is to support needs of travelers by storing in a central repository semantically demarcated data (data gathering rather then indexing [12]). In this note we will discuss how the content personalization aspects of our design by focusing on how user profiles are created, managed and used. The most important contribution of this work is to go beyond the standard work on personalization, presented for instance in [8, 30, 31, 32, 33, 34, 36], where the exact form of data to be operated on is either ignored or is radically different from that which we have proposed. We proceed as follows: in Section 2 we present the general architecture of the system and briefly sketch its functionality. Section 3 characterizes the travel-related data that the system manipulates and describes how this data is utilized in the context of user-system interactions (Section 4). In Section 5 we outline a basic user profile and methods for creating and updating profiles. Finally, in Section 6 we discuss how the user profile is actually employed in the system. We conclude the paper with description of current activities. 2. System architecture The helicopter view of the architecture of the proposed system is depicted in Figure 1. Let us briefly summarize