... this paper, we present a new type of recommendation system that carries out a personalized dialogue with the user. This system -- the Adaptive Place Advisor -- treats item selection as an interactive, conversational process, with the program inquiring about item attributes and the user responding. The system incorporates a user model that contains item, attribute, and value preferences, which it updates during each conversation and maintains across sessions. The Place Advisor uses both the conversational context and the user model to retrieve candidate items from a case base. The system then continues to ask questions, using personalized heuristics to select which attribute to ask about next, presenting complete items to the user only when a few remain. We report experimental results demonstrating the effectiveness of user modeling in reducing the time and number of interactions required to find a satisfactory item

In an experimental environment, we simulated the situation of a user who gives speech input to a system while walking through an airport. The time pressure on the subjects and the requirement to navigate while speaking were manipulated orthogonally. Each of the 32 subjects generated 80 utterances, which were coded semi-automatically with respect to a wide range of features, such as filled pauses. The experiment yielded new results concerning the effects of time pressure and cognitive load on speech. To see whether a system can automatically identify these conditions on the basis of speech input, we had this task performed for each subject by a Bayesian network that had been learned on the basis of the experimental data for the other subjects. The results shed light on the conditions that determine the accuracy of such recognition. 1 Background and Issues This paper is an experimental follow-up to the UM99 paper by Berthold and Jameson ([2]). Those authors argued the follo...

This article summarizes research on several interrelated general issues that can arise in the design and development of user modeling systems: the learning and subsequent adaptation of general user models on the basis of empirical data; the modeling of temporally variable properties of users, in particular time pressure and cognitive load; and the user-adaptive planning of interactions under uncertainty. The methods and results are integrated and illustrated with a prototype of a mobile assistance system for travelers in an airport.

This paper presents some motivations for using highly personalised speech interfaces. In particular, the focus is on the requirements for adaptation in mobile environments. Furthermore, SesaME, a framework for personalised and adaptive speech interfaces is described. SesaME supports a multi-domain approach and eventbased, asynchronous dialogue management.

This paper addresses the primary workshop question on "how to apply machine learning techniques to acquire and continuously adapt user models" for the particular representation of user models as Bayesian networks. On the basis of an integrative framework for learning Bayesian networks for user modeling and user-adaptive systems, respectively, we discuss some of the methods we developed along these lines, in the light of the questions that are to be discussed during the workshop. As this paper is intended to give an overview of our research, it omits some technical details, that can be found in related publications.

We discuss issues that arise when applying techniques for the learning of Bayesian networks in the user modeling context. We address the problem of sparse data that is often present in user modeling and show how we try to cope with it by introducing available a-priori knowledge into the learning procedures. Particularly, we present initial results concerning the learning of the structural part of a Bayesian network user model. 1

Abstract. User behaviors on a system vary not only among individuals but also within the same user when he/she gains experience on the system. We empirically investigated how individual users changed their behaviors on the basis of long-term data, which were collected by our telephone-based spoken dialogue system deployed for the open public over 34 months. The system was repeatedly used by citizens, who were each identified by their phone numbers. We conducted an experiment by using these data and showed that prediction accuracy of utteranceunderstanding errors improved when the temporal change was taken into consideration. This result showed that modeling temporally changing user behaviors was helpful in improving the performance of spoken dialogue systems.

Abstract. Context-aware applications and middleware platforms are evolving into major driving factors for pervasive systems. The ability to also make accurate assumptions about future contexts further enables such systems to proactively adapt to upcoming situations. However, the provision of a reusable system component to facilitate the development of such future-context-aware applications is still challenging- as it requires to be generic but, at the same time, as efficient and accurate as possible. To address these requirements, this paper presents the approach of Structured Context Prediction which constitutes a framework to facilitate the application of existing prediction methods. It allows application developers to integrate domain-specific knowledge by creating a customized prediction model at design time and to select, implement and combine prediction methods for the intended purpose. Feasibility is evaluated by applying a prototype system component to two mobile application scenarios, showing that both high accuracy and efficiency are possible. 1

Abstract. One of the central questions addressed in the project READY was that of how a system can automatically recognize situationally determined resource limitations of its user—in particular, time pressure and cognitive load. This chapter summarizes most of the work done in READY on this topic, presenting as well some previously unpublished results. We first consider why on-line recognition or resource limitations can be useful by discussing the ways in which a system might adapt its behavior to perceived resource limitations. We then summarize a number of approaches to the recognition problem that have been taken in READY and other projects, before focusing on one particular approach: the analysis of features of a user’s speech. In each of two similarly structured experiments, we created four experimental conditions that varied in terms of whether the user was (a) required to produce spoken utterances quickly or not; and (b) navigating within a simulated airport terminal or standing still. In the second experiment, additional distraction was caused by continuous loudspeaker announcements. The speech produced by the experimental subjects (32 in each experiment) was coded in terms of 7 variables. We report on the extent to which each of these variables was influenced by the subjects ’ resource limitations. We also trained dynamic Bayesian networks on the resulting data in order to see how well the information in the users ’ speech could serve as evidence as to which condition the user had been in. The results yield information about the accuracy that can be attained in this way and about the diagnostic value of some specific features of speech. 1