Automatic dialogue systems used, for instance, in call centers, should be able to determine in a critical phase of the dialogue––indicated by the customers vocal expression of anger/irritation––when it is better to pass over to a human operator. At a first glance, this does not seem to be a complicated task: It is reported in the literature that emotions can be told apart quite reliably on the basis of prosodic features. However, these results are achieved most of the time in a laboratory setting, with experienced speakers (actors), and with elicited, controlled speech. We compare classification results obtained with the same feature set for elicited speech and for a Wizard-of-Oz scenario, where users believe that they are really communicating with an automatic dialogue system. It turns out that the closer we get to a realistic scenario, the less reliable is prosody as an indicator of the speakersÕ emotional state. As a consequence, we propose to change the target such that we cease looking for traces of particular emotions in the usersÕ speech, but instead look for indicators of TROUBLE INCOMMUNICATION. INCOMMUNICATION For this reason, we propose the module Monitoring of User State [especially of] Emotion (MOUSE MOUSE) in which a prosodic classifier is combined with other knowledge sources, such as conversationally peculiar linguistic behavior, for example, the use of repetitions. For this module, preliminary exper-imental results are reported showing a more adequate modelling of TROUBLE INCOMMUNICATION.

Spoken dialogue system performance can vary widely for different users, as well for the same user during different dialogues. This paper presents the design and evaluation of an adaptive version of TOOT, a spoken dialogue system for retrieving online train schedules. Adaptive TOOT predicts whether a user is having speech recognition problems as a particular dialogue progresses, and automatically adapts its dialogue strategies based on its predictions. An empirical evaluation of the system demonstrates the utility of the approach. Introduction Most spoken dialogue systems do not try to improve performance by dynamically adapting the system's dialogue behaviors to an individual user during the course of a particular dialogue. But the performance of a spoken dialogue system can vary significantly for different users and even for the same user across dialogues. This paper presents the design and experimental evaluation of a spoken dialogue system that predicts and responds to pr...

While it has recently become possible to build spoken dialogue systems that interact with users in real-time in a range of domains, systems that support conversational natural language are still subject to a large number of spoken language understanding (SLU) errors. Endowing such systems with the ability to reliably distinguish SLU errors from correctly understood utterances might allow them to correct some errors automatically or to interact with users to repair them, thereby improving the system's overall performance. We report experiments on learning to automatically distinguish SLU errors in 11,787 spoken utterances collected in a field trial of AT&T's How May I Help You system interacting with live customer traffic. We apply the automatic classifier RIPPER (Cohen 96) to train an SLU classifier using features that are automatically obtainable in real-time. The classifer achieves 86% accuracy on this task, an improvement of 23% over the majority class baseline....

Models of computer users that are learned on the basis of data can make use of two types of information: data about users in general and data about the current individual user. Focusing on user models that take the form of Bayesian networks, we compare four types of model that represent different ways of combining these two types of data. Models of the four types are applied to the data of an experiment, and they are evaluated according to theoretical, empirical, and practical criteria. One of the model types is a new variant of the AHUGIN method for adapting the probabilities of a Bayesian network while it is being used: Differential adaptation is a principled way of determining the speed with which each aspect of a network is adapted to an individual user. 1

sources and services from any phone. However, current spoken dialogue systems are deficient in their strategies for preventing, identifying and repairing problems that arise in the conversation. This paper reports results on automatically training a Problematic Dialogue Predictor to predict problematic human-computer dialogues using a corpus of 4692 dialogues collected with the How May I Help You spoken dialogue system. The Problematic Dialogue Predictor can be immediately applied to the system's decision of whether to transfer the call to a human customer care agent, or be used as a cue to the system's dialogue manager to modify its behavior to repair problems, and even perhaps, to prevent them. We show that a Problematic Dialogue Predictor using automaticallyobtainable features from the first two exchanges in the dialogue can predict problematic dialogues 13.2% more accurately than the baseline.

This paper describes the user expertise model in AthosMail, a mobile, speech-based e-mail system. The model encodes the system's assumptions about the user expertise, and gives recommendations on how the system should respond depending on the assumed competence levels of the user. The recommendations are realized as three types of explicitness in the system responses. The system monitors the user's competence with the help of parameters that describe e.g. the success of the user's interaction with the system. The model consists of an online and an offline version, the former taking care of the expertise level changes during the same session, the latter modelling the overall user expertise as a function of time and repeated interactions.

This article focuses on the analysis and prediction of corrections, defined as turns where a user tries to correct a prior error made by a spoken dialogue system. We describe our labeling procedure of various corrections types and statistical analyses of their features in a corpus collected from a train information spoken dialogue system. We then present results of machinelearning experiments designed to identify user corrections of speech recognition errors. We investigate the predictive power of features automatically computable from the prosody of the turn, the speech recognition process, experimental conditions, and the dialogue history. Our best-performing features reduce classification error from baselines of 25.70–28.99 % to 15.72%. 1.

This thesis addresses the cognitive basis of syntactic adaptation, which biases speakers to repeat their own syntactic constructions and those of their conversational partners. I address two types of syntactic adaptation: short-term priming and longterm adaptation. I develop two metrics for syntactic adaptation within a speaker and between speakers in dialogue: one for short-term priming effects that decay quickly, and one for long-term adaptation over the course of a dialogue. Both methods estimate adaptation in large datasets consisting of transcribed human-human dialogue annotated with syntactic information. Two such corpora in English are used: Switchboard, a collection of spontaneous phone conversation, and HCRC Map Task, a set of task-oriented dialogues in which participants describe routes on a map to one another. I find both priming and long-term adaptation in both corpora, confirming well-known experimental results (e.g., Bock, 1986b). I extend prior work by showing that syntactic priming effects not only apply to selected syntactic constructions that are alternative realizations of the same semantics, but still hold when a broad

Organizations are increasingly turning to spoken dialog systems for automated call routing to reduce call center costs. To maintain quality service even in cases of failure, these systems often resort to ad-hoc rules for dispatching calls to a human operator. We present a principled procedure for determining when callers should be transferred to operators based on a cost-benefit analysis. The procedure integrates models that predict when a call is likely to fail using spoken dialog features with queuing models of call center volume and service time. We evaluate how the procedure would have performed on cases drawn from logs of interactions with a legacy spoken dialog system. 1

Abstract not found