This paper describes a way of using intonation and dialogue context to improve the performance of an automatic speech recognition (ASR) system. Our experiments were run on the DCIEM Maptask corpus, a corpus of spontaneous task-oriented dialogue speech. This corpus has been tagged according to a dialogue analysis scheme that assigns each utterance to one of 12 "move types", such as "acknowledge", "query-yes/no" or "instruct". Most asr systems use a bigram language model to constrain the possible sequences of words that might be recognised. Here we use a separate bigram language model for each move type. We show that when the "correct" move-specific language model is used for each utterance in the test set, the word error rate of the recogniser drops. Of course

The thesis describes a logical formalization of natural-language database interfacing. We assume the existence of a "natural language engine" capable of mediating between surface linguistic string and their representations as "literal" logical forms: the focus of interest will be the question of relating "literal" logical forms to representations in terms of primitives meaningful to the underlying database engine. We begin by describing the nature of the problem, and show how a variety of interface functionalities can be considered as instances of a type of formal inference task which we call "Abductive Equivalential Translation" (AET); functionalities which can be reduced to this form include answering questions, responding to commands, reasoning about the completeness of answers, answering meta-questions of type "Do you know...", and generating assertions and questions. In each case, a "linguistic domain theory" (LDT) \Gamma and an input formula F are given, and the goal is to constr...

The paper uses a simple and abstract characterization of dialogue in terms of mental state changes of dialogue participants to raise three fundamental questions for any theory of dialogue. It goes on to discuss currently popular accounts of dialogue with respect to these three questions. Next, the notion of `conversational game' is revisited within a probabilistic and decision theoretic framework, and it is argued that such an interpretation is plausible both intuitively and as the basis for computational implementation. An illustrated sketch of a proposed implementation using Bayesian networks is described. Three Questions for Dialogue A simple, rather abstract description of a canonical dialogue is that it consists of a sequence of utterances with a corresponding sequence of mental states of the participants in the dialogue. Person A has a sequence of mental states SA1 . . . SAn+1 and person B also has a sequence SB1 . . . SBn+1 . Connecting these two sequences is a third sequence, ...

Many of the kinds of language model used in speech understanding suffer from imperfect modeling of intra-sentential contextual influences. I argue that this problem can be addressed by clustering the sentences in a training corpus automatically into subcorpora on the criterion of entropy reduction, and calculating separate language model parameters for each cluster. This kind of clustering offers a way to represent impor- tant contextual effects and can therefore significantly improve the performance of a model. It also offers a reasonably automatic means to gather evidence on whether a more complex, context-sensitive model using the same general kind of linguistic information is likely to reward the effort that would be required to develop it: if clustering improves the performance of a model, this proves the existence of further context dependencies, not exploited by the unclustered model. As evidence for these claims, I present results showing that clustering improves some models but not others for the ATIS domain. These results are consistent with other findings for such models, suggesting that the existence or otherwise of an improvement brought about by clustering is indeed a good pointer to whether it is worth developing further the unclustered model.

We look at the effect of using high level discourse knowledge in dialogue act type detection. We also look at ways this knowledge can be used for improving language modelling and intonation modelling of utterance types. We find a significant improvement of predictability of dialogue models using higher level discourse knowledge. 1. INTRODUCTION This paper describes a method of dialogue act recognition that takes advantage of regularities in discourse at various levels from prosodic features to goal oriented dialogue information. Pragmatic theory (Levinson [7]) suggests that conversation follows a script and that sequences of dialogue acts are not random. For example, a query followed by a response, followed by an acknowledgement is more likely than three acknowledgements in succession. Dialog act identification is an important part of dialog systems such as Allen et al. [1], Lewin et al. [8]. It can also be used in automatic speech recognition systems to improve word error rate, Shr...

this paper is to investigate and compare coding schemas for dialogue moves

We discuss the treatment of ellipsis in a spoken language route planning enquiry service which uses the Core Language Engine (CLE) as its linguistic processor. We show how use of the CLE allows us to separate the interpretation of ellipsis in a dialogue context from the more general issue of dialogue management in a dialogue context and, especially, to factor out the linguistic influences on such interpretation and place them where they belong - in the linguistic processor. The route planning application was developed with a Wizard of Oz corpus to help guide it and we discuss the application of the CLE to some particularly interesting and potentially troublesome data from that corpus. 1. INTRODUCTION A spoken language route planning enquiry service has been developed [9] which uses the Core Language Engine (CLE) [1, 2] as its linguistic processor. The CLE is a wide-coverage general purpose natural language processing system with facilities for reasoning and understanding in context. ...

Acknowledgements I would like to take this opportunity to thank all the people who have helped me with my research and thus made this work possible. First of all, I am very grateful to my advisor, Alexander Rudnicky, for his valuable guidance and support throughout the course of this work and my graduate school career. His vision and understanding in scientific research teach me all the elements I would need to know when conducting high quality research on my own. I am also sincerely grateful to my committee members, William Cohen, Carolyn Penstein Rosé, and Gokhan Tür, for their valuable suggestions and contributions to this research. Carolyn Penstein Rosé in particular gave me a great deal of feedback on the thesis writing. While at CMU, I have been lucky enough to be a part of a friendly and very helpful research community. I would like to thank my fellow research group members and the members of the Dialogs on Dialogs reading group, Dan Bohus, Antoine Raux, Mihai Rotaru, Banerjee Satanjeev, Stefanie Tomko, and many other people that I would not be possible to list all their names here, for numerous intellectual discussions and their

The candidate confirms that the work submitted is her own and that appropriate credit has been given where reference has been made to the work of others. This copy is supplied on the understanding that it is copyright material and that no quotation from the thesis may be published without proper acknowledgementFor my father, David Jorge Schiffrin, Caminante, son tus huellas El camino y nada más; Caminante, no hay camino, Se hace camino al andar. Traveller, your footprints Are the road and nothing else; Traveller, there is no road, You make the road by walking. Al andar se hace camino Y al volver la vista atrás Se ve la senda que nunca

Abstract not found