Challenges in Building Highly Interactive Dialogue Systems

Ward, Nigel; DeVault, David

doi:10.1609/aimag.v37i4.2687

Cited by 28 publications

(24 citation statements)

References 71 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, the popular GeMAPS set, designed for paralinguistic inference from prosody, treats energy and pitch features as separate streams, meaning that it is unable to capture alignment phenomena at all [43]. However, the current study has identified disalignment-related functions that include aspects that are very widely relevant -including for better second language teaching to foster communicative effectiveness, improving speech synthesis to more effectively convey dialog intentions, improving dialog act recognition, and better tracking of user intentions and attitudes in dialog systems [44]. Thus future work aiming to better understand peak alignment phenomenon is likely to have not only theoretical importance but also great practical value.…”

Section: Summary and Open Questionsmentioning

confidence: 92%

A Corpus-Based Exploration of the Functions of Disaligned Pitch Peaks in American English Dialog

Ward¹

2018

Speech Prosody 2018

View full text Add to dashboard Cite

The exact positioning of pitch peaks often has communicative significance, but the meanings and functions this conveys have never been systematically studied. This paper reports an exploration of one basic aspect of this in one language. The phenomenon is that of "disaligned pitch peaks," that is, peaks which, contrary to the usual tendency, are not aligned with a strong energy peak. The language is American English, as used in naturally-occurring two-person interactions. To find examples, I developed a model to automatically estimate the extent to which a speech signal exhibits a strong pitch peak that is not aligned with an energy peak. Examination of examples revealed many associated pragmatic functions, including suggesting, grounding, agreeing with reservations, implying, and expressing liking, most of which have not been previously noted.

show abstract

Section: Summary and Open Questionsmentioning

confidence: 92%

A Corpus-Based Exploration of the Functions of Disaligned Pitch Peaks in American English Dialog

Ward¹

2018

Speech Prosody 2018

View full text Add to dashboard Cite

show abstract

“…They could be used for predicting a player's likely future actions from his or her prosody, or to help a robot player choose his actions based on a human player's utterance prosody. More generally, models of this kind could be useful for creating systems better able to coordinate their actions with human partners [42], for situated-action domains and beyond.…”

Section: Summary Directions For Future Work and Potential Valuementioning

confidence: 99%

Action-coordinating prosody

Ward¹,

Abu

2016

Speech Prosody 2016

View full text Add to dashboard Cite

This paper is an initial exploration of how prosody helps coordinate action, based on examination of speech and motion in a two-player maze game where the players run and jump to avoid obstacles, and coordinate movements to solve problems. We use an unsupervised method, Principal Component Analysis applied to a large set of time-spread features, to discover patterns of behavior involving both prosodic features and game actions. These patterns include prosodic constructions involved in assessing, planning, inhibiting, cuing, and synchronizing actions.

show abstract

“…Dialogue management for spoken dialogue systems is a challenging research domain due in part to difficulties arising from limited resources, the imperfection of technologies on which dialogue management is dependent, and of course the complexities of natural human conversation (Glass, 1999;Ward and DeVault, 2015). Within a conventional dialogue manager, an explicit dialogue state tracker is a key component that attempts to track both interlocutors' contributions to the exchange.…”

Section: Introductionmentioning

confidence: 99%

Capturing Dialogue State Variable Dependencies with an Energy-based Neural Dialogue State Tracker

Trinh¹,

Ross

Kelleher³

2019

Proceedings of the 20th Annual SIGdial Meeting on Discourse and Dialogue

View full text Add to dashboard Cite

Dialogue state tracking requires the population and maintenance of a multi-slot frame representation of the dialogue state. Frequently, dialogue state tracking systems assume independence between slot values within a frame. In this paper we argue that treating the prediction of each slot value as an independent prediction task may ignore important associations between the slot values, and, consequently, we argue that treating dialogue state tracking as a structured prediction problem can help to improve dialogue state tracking performance. To support this argument, the research presented in this paper is structured into three stages: (i) analyzing variable dependencies in dialogue data; (ii) applying an energy-based methodology to model dialogue state tracking as a structured prediction task; and (iii) evaluating the impact of inter-slot relationships on model performance. Overall, we demonstrate that modelling the associations between target slots with an energy-based formalism improves dialogue state tracking performance in a number of ways.

show abstract

Challenges in Building Highly Interactive Dialogue Systems

Cited by 28 publications

References 71 publications

A Corpus-Based Exploration of the Functions of Disaligned Pitch Peaks in American English Dialog

A Corpus-Based Exploration of the Functions of Disaligned Pitch Peaks in American English Dialog

Action-coordinating prosody

Capturing Dialogue State Variable Dependencies with an Energy-based Neural Dialogue State Tracker

Contact Info

Product

Resources

About