Speaking Styles and Interaction Styles
Motivation
Many core speech technologies are now highly tuned and rival or
exceed human performance. However, the user experience is often
lacking, in part because speech synthesizers, speech recognizers,
dialog managers, and other components are still inflexible in their
expectations of human behavior. Moreover, considerations of style are
becoming more important as we seek to deploy variations of one basic
system across domains and genres. Thus more attention to interaction
style variation and speaking style variation may help us improve the
value for users.
Recently work in these areas has been rapidly advancing. We now
know more about many types of style, including expressive styles,
styles relating to social roles, and styles relating to stance and
interpersonal dynamics. There are now many techniques for style in
speech synthesis and voice conversion, style transfer in text and
speech. The connections of style to genre, priming,
accommodation/entrainment, lexical choice, social identity, dialog
activity modeling, and next dialog move selection, have been
increasingly examined.
However, work on style has been fragmented. There has been little
communication between those working on style as it relates to
various core technologies, various phenomena, various
applications needs, and various linguistic questions.
Accordingly, the aim of this interdisciplinary special session is
to bring together researchers who work on different aspects of style
for spoken dialog, who work on different tasks, or who approach style
from different perspectives. We expect this will foster
cross-fertilization, broader understanding of the goals and
challenges, and identification of shared priorities for further
research.
Format
We envisage two oral sessions, with the last hour of
the second session devoted to a panel discussion, plus one poster
session.
Call for Papers
We welcome submissions on all aspects of style in speech and
interaction. All papers must have an empirical component and must
make a significant contribution to speech science or speech
technology. We further encourage authors to also elaborate on how
their work contributes to or relates to "the big picture" of style in
spoken dialog. We also welcome papers whose motivations,
contributions, or implications relate to more than one traditional
Interspeech area, or which highlight issues not commonly addressed at
Interspeech.
Questions of interest include:
How can we relate defined/explainable/controllable styles and
learned/latent styles?
What aspects of style matter most for spoken language systems
and for other applications?
What general principles and modeling techniques underlie style
transfer across speech synthesis, language generation, and interactive
behaviors?
How do the dimensions of style differ across domains, genres, and
languages?
Can we approach a universal style dictionary or a universal space
of styles?
What commonalities are there among expressive styles, social
styles, and interpersonal styles?
How do styles relate to personality, gender, social identity, and dialect?
How can style modeling support influence detection and other
classification tasks?
Which of the observations and aspirations of classic work on style
(by Tannen, Biber, Pennebaker, and others) are not yet treated by
computational models, and how can we bridge the gaps?
To what extent are style inventories and methods developed for
monologue data sufficient to handle styles as they occur in dialog?
To what extent are findings based on controlled data relevant to
style in the wild?
What range of interaction styles do users consider acceptable for
robots and other interactive systems?
How does style variation operate across different timescales, from
the utterance level, through transient stances and interpersonal
attitudes, to the dialog and installation levels?
How do style choices serve both the overall goals of the
interaction and the local goals and dialog activities?
How do less-studied aspects, such as voice quality, phonetic
variation, turn timing, and gestures, contribute to the realization of
styles?
Which aspects of style variation are most basic, perceptually or
developmentally?
How can models of styles and their realizations in one domain
transfer to other domains?
How can we model human perceptions and misperceptions of style?
How can style matching contribute to user acceptance, decrease
cognitive load, or improve rapport?
How can knowledge of style differences improve the teaching,
learning, and assessment of language and communications skills?
Submission
Please use
the standard Interspeech style and format, and submit papers through
the standard Interspeech process.
Organizers
Nigel Ward, University of Texas at El Paso
Kallirroi Georgila, University of Southern California
Scientific Committee
Yang Gao , Amazon AWS AI
Mark Hasegawa-Johnson, University of Illinois
Koji Inoue, Kyoto University
Simon King ,
University of Edinburgh
Rivka Levitan, City University of New York
Katherine Metcalf, Apple
Eva Szekely, KTH Royal Institute for Technology
Pol van Rijn, Max Planck Institute for Empirical Aesthetics
Rafael Valle , NVIDIA
