The past decade has seen a revolution in the field of spoken dialogue systems. As in other areas of Computer Science and Artificial Intelligence, data-driven methods are now being used to drive new methodologies for system development and evaluation. This book is a unique contribution to that ongoing change. A new methodology for developing spoken dialogue systems is described in detail. The journey starts and ends with human behaviour in interaction, and explores methods for learning from the data, for building simulation environments for training and testing systems, and for evaluating the results. The detailed material covers: Spoken and Multimodal dialogue systems, Wizard-of-Oz data collection, User Simulation methods, Reinforcement Learning, and Evaluation methodologies. The book is a research guide for students and researchers with a background in Computer Science, AI, or Machine Learning. It navigates through a detailed case study in data-driven methods for development and evaluation of spoken dialogue systems. Common challenges associated with this approach are discussed and example solutions are provided. This work provides insights, lessons, and inspiration for future research and development – not only for spoken dialogue systems in particular, but for data-driven approaches to human-machine interaction in general.

This document proposes to learn the behaviour of the dialogue manager of a spoken dialogue system from a set of rated dialogues. This learning is performed through reinforcement learning. Our method does not require the definition of a representation of the state space nor a reward function. These two high-level parameters are learnt from the corpus of rated dialogues. It is shown that the spoken dialogue designer can optimise dialogue management by simply defining the dialogue logic and a criterion to maximise (e.g user satisfaction). The methodology suggested in this thesis first considers the dialogue parameters that are necessary to compute a representation of the state space relevant for the criterion to be maximized. For instance, if the chosen criterion is user satisfaction then it is important to account for parameters such as dialogue duration and the average speech recognition confidence score. The state space is represented as a sparse distributed memory. The Genetic Sparse Distributed Memory for Reinforcement Learning (GSDMRL) accommodates many dialogue parameters and selects the parameters which are the most important for learning through genetic evolution. The resulting state space and the policy learnt on it are easily interpretable by the system designer. Secondly, the rated dialogues are used to learn a reward function which teaches the system to optimise the criterion. Two algorithms, reward shaping and distance minimisation are proposed to learn the reward function. These two algorithms consider the criterion to be the return for the entire dialogue. These functions are discussed and compared on simulated dialogues and it is shown that the resulting functions enable faster learning than using the criterion directly as the final reward. A spoken dialogue system for appointment scheduling was designed during this thesis, based on previous systems, and a corpus of rated dialogues with this system were collected. This corpus illustrates the scaling capability of the state space representation and is a good example of an industrial spoken dialogue system upon which the methodology could be applied.

Data driven methods have long been used in Automatic Speech Recognition (ASR) and Text-To-Speech (TTS) synthesis and have more recently been introduced for dialogue management, spoken language understanding, and Natural Language Generation. Machine learning is now present “end-to-end” in Spoken Dialogue Systems (SDS). However, these techniques require data collection and annotation campaigns, which can be time-consuming and expensive, as well as dataset expansion by simulation. In this book, we provide an overview of the current state of the field and of recent advances, with a specific focus on adaptivity.

In Monitoring Adaptive Spoken Dialog Systems, authors Alexander Schmitt and Wolfgang Minker investigate statistical approaches that allow for recognition of negative dialog patterns in Spoken Dialog Systems (SDS). The presented stochastic methods allow a flexible, portable and accurate use. Beginning with the foundations of machine learning and pattern recognition, this monograph examines how frequently users show negative emotions in spoken dialog systems and develop novel approaches to speech-based emotion recognition using hybrid approach to model emotions. The authors make use of statistical methods based on acoustic, linguistic and contextual features to examine the relationship between the interaction flow and the occurrence of emotions using non-acted recordings several thousand real users from commercial and non-commercial SDS. Additionally, the authors present novel statistical methods that spot problems within a dialog based on interaction patterns. The approaches enable future SDS to offer more natural and robust interactions. This work provides insights, lessons and inspiration for future research and development, not only for spoken dialog systems, but for data-driven approaches to human-machine interaction in general.

This book covers state-of-the-art topics on the practical implementation of Spoken Dialog Systems and intelligent assistants in everyday applications. It presents scientific achievements in language processing that result in the development of successful applications and addresses general issues regarding the advances in Spoken Dialog Systems with applications in robotics, knowledge access and communication. Emphasis is placed on the following topics: speaker/language recognition, user modeling / simulation, evaluation of dialog system, multi-modality / emotion recognition from speech, speech data mining, language resource and databases, machine learning for spoken dialog systems and educational and healthcare applications.

This book addresses the problems of spoken dialogue system design and especially automatic learning of optimal strategies for man-machine dialogues. Besides the description of the learning methods, this text proposes a framework for realistic simulation of human-machine dialogues based on probabilistic techniques, which allows automatic evaluation and unsupervised learning of dialogue strategies. This framework relies on stochastic modelling of modules composing spoken dialogue systems as well as on user modelling. Special care has been taken to build models that can either be hand-tuned or learned from generic data.