Policy optimisation of POMDP-based dialogue systems without state space compression

The partially observable Markov decision process (POMDP) has been proposed as a dialogue model that enables automatic improvement of the dialogue policy and robustness to speech understanding errors. It requires, however, a large number of dialogues to train the dialogue policy. Gaussian processes (GP) have recently been applied to POMDP dialogue management optimisation showing an ability to substantially increase the speed of learning. Here, we investigate this further using the Bayesian Update of Dialogue State dialogue manager. We show that it is possible to apply Gaussian processes directly to the belief state, removing the need for a parametric policy representation. In addition, the resulting policy learns significantly faster while maintaining operational performance. © 2012 IEEE.

Gaussian processes for POMDP-based dialogue manager optimization

A partially observable Markov decision process (POMDP) has been proposed as a dialog model that enables automatic optimization of the dialog policy and provides robustness to speech understanding errors. Various approximations allow such a model to be used for building real-world dialog systems. However, they require a large number of dialogs to train the dialog policy and hence they typically rely on the availability of a user simulator. They also require significant designer effort to hand-craft the policy representation. We investigate the use of Gaussian processes (GPs) in policy modeling to overcome these problems. We show that GP policy optimization can be implemented for a real world POMDP dialog manager, and in particular: 1) we examine different formulations of a GP policy to minimize variability in the learning process; 2) we find that the use of GP increases the learning rate by an order of magnitude thereby allowing learning by direct interaction with human users; and 3) we demonstrate that designer effort can be substantially reduced by basing the policy directly on the full belief space thereby avoiding ad hoc feature space modeling. Overall, the GP approach represents an important step forward towards fully automatic dialog policy optimization in real world systems. © 2013 IEEE.

À propos de The Practice of Liberal Pluralism de William Galston : un dialogue avec l’auteur

Document de travail

L'altérité selon Lévinas et Ricoeur comme prémisse éthique au dialogue judéo-chrétien

L‟objectif de mon mémoire se concentre sur la notion d‟altérité émanant des philosophies d‟Emmanuel Lévinas et de Paul Ricoeur ; je m‟intéresse plus précisément au concept clé d‟éthique et de savoir en quoi enrichit-elle le dialogue judéo-chrétien. Le point initial de ma réflexion est l‟herméneutique biblique, qu‟Emmanuel Lévinas et Paul Ricoeur articulent, d‟après moi, différemment selon leurs héritages religieux respectifs à savoir juif et chrétien. Néanmoins, la signification éthique des Textes Sacrés perdure pour chacun d‟eux comme lieu commun même si la signification leur est différente et propre à leurs traditions religieuses. Ainsi, dans ce mémoire l‟altérité développée par Lévinas, talmudiste reconnu, sera comparée avec la pensée de Ricoeur dont la conception est davantage chrétienne, en référence à son travail exégétique. Quand bien même Lévinas et Ricoeur ont tenu à distinguer leurs philosophies de leurs théologies, l‟hypothèse de départ prend une liberté herméneutique qui oscille souvent entre philosophie et théologie et qui tend à retracer au mieux l‟altérité et son lien intrinsèque avec l‟éthique. Cette lecture comparatiste m‟amènera donc à penser et à intégrer l‟altérité comme une prémisse éthique au dialogue judéo-chrétien. Mon travail en sciences des religions qui prend racine depuis l‟herméneutique même, s‟oriente vers une perspective éthique et dialogique et c‟est cette visée de médiation interreligieuse qui lui confère une appartenance à cette discipline.

Educating integrated catalysts: transforming business schools toward ethics and sustainability

Joining the sharpening critique of conventional University-based business school education, we argue that educating integrated catalysts is necessary to meet current sustainability challenges. The key feature of moving toward the integration required at the individual level is focusing on developing students' capacity for moral and cognitive maturity. Practically, this makes the practice of genuine dialogue focal as core interpersonal method for educating management students. In supporting such education, business schools must however first transform themselves. Acting as transformative social enterprises, they can demonstrate being a part in critically questioning and improving the impact and relevance of management on the flourishing of wider society and the practice of an ethically oriented economy. We offer practical suggestions and implications for future business education reform.

On the dynamic adaptation of language models based on dialogue information

We present an approach to adapt dynamically the language models (LMs) used by a speech recognizer that is part of a spoken dialogue system. We have developed a grammar generation strategy that automatically adapts the LMs using the semantic information that the user provides (represented as dialogue concepts), together with the information regarding the intentions of the speaker (inferred by the dialogue manager, and represented as dialogue goals). We carry out the adaptation as a linear interpolation between a background LM, and one or more of the LMs associated to the dialogue elements (concepts or goals) addressed by the user. The interpolation weights between those models are automatically estimated on each dialogue turn, using measures such as the posterior probabilities of concepts and goals, estimated as part of the inference procedure to determine the actions to be carried out. We propose two approaches to handle the LMs related to concepts and goals. Whereas in the first one we estimate a LM for each one of them, in the second one we apply several clustering strategies to group together those elements that share some common properties, and estimate a LM for each cluster. Our evaluation shows how the system can estimate a dynamic model adapted to each dialogue turn, which helps to improve the performance of the speech recognition (up to a 14.82% of relative improvement), which leads to an improvement in both the language understanding and the dialogue management tasks.

A Dialogue between Treatment Integrity Research and Design-Based Implementation Research

Thesis (Master's)--University of Washington, 2016-06

Document retrieval based on a cognitive model of dialogue

Owing to the rise in the volume of literature, problems arise in the retrieval of required information. Various retrieval strategies have been proposed, but most of that are not flexible enough for their users. Specifically, most of these systems assume that users know exactly what they are looking for before approaching the system, and that users are able to precisely express their information needs according to l aid- down specifications. There has, however, been described a retrieval program THOMAS which aims at satisfying incompletely- defined user needs through a man- machine dialogue which does not require any rigid queries. Unlike most systems, Thomas attempts to satisfy the user's needs from a model which it builds of the user's area of interest. This model is a subset of the program's "world model" - a database in the form of a network where the nodes represent concepts since various concepts have various degrees of similarities and associations, this thesis contends that instead of models which assume equal levels of similarities between concepts, the links between the concepts should have values assigned to them to indicate the degree of similarity between the concepts. Furthermore, the world model of the system should be structured such that concepts which are related to one another be clustered together, so that a user- interaction would involve only the relevant clusters rather than the entire database such clusters being determined by the system, not the user. This thesis also attempts to link the design work with the current notion in psychology centred on the use of the computer to simulate human cognitive processes. In this case, an attempt has been made to model a dialogue between two people - the information seeker and the information expert. The system, called Thomas-II, has been implemented and found to require less effort from the user than Thomas.