"Letizia Alvarez de Toledo has observed that this vast library is useless: rigorously speaking, a single volume ■ would be sufficient, a volume of ordinary format, printed in nine or ten point type, containing an infinite

This paper presents an overview of the field of recommender systems and describes the current generation of recommendation methods that are usually classified into the following three main categories: content-based, collaborative, and hybrid recommendation approaches. This paper also describes various limitations of current recommendation methods and discusses possible extensions that can improve recommendation capabilities and make recommender systems applicable to an even broader range of applications. These extensions include, among others, an improvement of understanding of users and items, incorporation of the contextual information into the recommendation process, support for multcriteria ratings, and a provision of more flexible and less intrusive types of recommendations.

Personal software assistants that help users with tasks like finding information, scheduling calendars, or managing work-flow will require significant customization to each individual user. For example, an assistant that helps schedule a particular user’s calendar will have to know that user’s scheduling preferences. This paper explores the potential of machine learning methods to automatically create and maintain such customized knowledge for personal software assistants. We describe the design of one particular learning assistant: a calendar manager, called CAP (Calendar APprentice), that learns user scheduling preferences from experience. Results are summarized from approximately five user-years of experience, during which CAP has learned an evolving set of several thousand rules that characterize the scheduling preferences of its users. Based on this experience, we suggest that machine learning methods may play an important role in future personal software assistants.

One category of researchers in artificial life is concerned with modeling and building so-called adaptive autonomous agents. Autonomous agents are systems that inhabit a dynamic, unpredictable environment in which they try to satisfy a set of time-dependent goals or motivations. Agents are said to be adaptive if they improve their competence at dealing with these goals based on experience. Autonomous agents constitute a new approach to the study of artificial intelligence (AI) which is highly inspired by biology, in particular ethology, the study of animal behavior. Research in autonomous agents has brought about a new wave of excitement into the field of AI. This paper reflects on the state of the art of this new approach.

Grouping people into clusters based on the items they have purchased allows accurate recommendations of new items for purchase: if you and I have liked many of the same movies, then I will probably enjoy other movies that you like. Recommending items based on similarity of interest (a.k.a. collaborative filtering) is attractive for many domains: books, CDs, movies, etc., but does not always work well. Because data are always sparse -- any given person has seen only a small fraction of all movies -- much more accurate predictions can be made by grouping people into clusters with similar movies and grouping movies into clusters which tend to be liked by the same people. Finding optimal clusters is tricky because the movie groups should be used to help determine the people groups and visa versa. We present a formal statistical model of collaborative filtering, and compare different algorithms for estimating the model parameters including variations of K-means clustering and Gibbs Sampling. This...

By combining both collaborative and content-based filtering systems, Fab may eliminate many of the weaknesses found in each approach. ONLINE READERS ARE IN NEED OF TOOLS TO HELP THEM COPE with the mass of content available on the World-Wide Web. In traditional media, readers are provided assistance in making selections. This includes both implicit assistance in the form of editorial oversight and explicit assistance in the form of recommendation services such as movie reviews and restaurant guides. The electronic medium offers new opportunities to create recommendation services, ones that adapt over time to track their evolving interests. Fab is such a recommendation system for the Web, and has been operational in several versions since December 1994. The problem of recommending items from some fixed database has been studied extensively, and two main paradigms have emerged. In content-based recommendation one tries to recommend items similar to those a given user has liked in the past, whereas in collaborative recommendation one identifies users whose tastes are similar to those of the given user and recommends items they have liked. Our approach in Fab has been to combine these two methods. Here, we explain how a hybrid system can incorporate the advantages of both methods while inheriting the disadvantages of neither. In addition to what one might call the “generic advantages ” inherent in any hybrid system, the particular design of the Fab architecture brings two additional benefits. First, two scaling problems common to all Web services are addressed—an increas-

An adaptive recommendation service seeks to adapt to its users, providing increasingly personalized recommendations over time. In this paper we introduce the "Fab" adaptive web page recommendation service. There has been much research on analyzing document content in order to improve recommendations or search results. More recently researchers have begun to explore how the similarities between users can be exploited to the same ends. The Fab system strikes a balance between these two approaches, taking advantage of the shared interests among users without losing the benefits of the representations provided by content analysis. Running since March 1996, it has been populated with a collection of agents for the collection and selection of web pages, whose interaction fosters emergent collaborative properties. In this paper we explain the design of the system architecture and report the results of our first experiment, evaluating recommendations provided to a group of test users. 1 Introd...

. A rapidly growing number of user and student modeling systems have employed numerical techniques for uncertainty management. The three major paradigms are those of Bayesian networks, the Dempster-Shafer theory of evidence, and fuzzy logic. In this overview, each of the first three main sections focuses on one of these paradigms. It first introduces the basic concepts by showing how they can be applied to a relatively simple user modeling problem. It then surveys systems that have applied techniques from the paradigm to user or student modeling, characterizing each system within a common framework. The final main section discusses several aspects of the usability of these techniques for user and student modeling, such as their knowledge engineering requirements, their need for computational resources, and the communicability of their results. Key words: numerical uncertainty management, Bayesian networks, Dempster-Shafer theory, fuzzy logic, user modeling, student modeling 1. Introdu...

We introduce a new genre of user interface applications that can migrate from one machine to another, taking their user interface and application contexts with them, and continue from where they left off. Such applications are not tied to one user or one machine, and can roam freely over the network, rendering service to a community of users, gathering human input and interacting with people. We envisage that this will support many new agent-based collaboration metaphors. The ability to migrate executing programs has applicability to mobile computing as well. Users can

A personalized information filtering system must specialize to current interests of the user and adapt as they change over time. It must also explore newer domains for potentially interesting information. A learning approach to building personalized information filtering systems is proposed. The system is designed as a collection of information filtering interface agents. Interface Agents are intelligent and autonomous computer programs which learn users' preferences and act on their behalf --- electronic personal assistants that automate tasks for the user. This thesis presents the basic framework for personalized information filtering agents, and describes an implementation, "Newt", built using the framework. Newt uses a keyword based filtering algorithm. The learning mechanisms used are relevance feedback and the genetic algorithm. The user interface is friendly and accessible to both naive as well as power users. Experimental