Stance-taking, the expression of opinions or attitudes, informs the process of negotiation, argumentation, and decision-making. While receiving significant attention in text materials in work on the related areas of subjectivity and sentiment analysis, the expression of stance in speech remains less explored. Prior analysis of the acoustics of stance-expression in conversational speech has identified some significant differences across dimensions of stance-related behavior. However, that analysis, as in much prior work, relied on simple functionals of pitch, energy, and duration, including maxima, minima, means, and ranges. In contrast, the current work focuses on exploiting measures that capture the dynamics of the pitch and energy contour. We employ features based on subband autocorrelation measures of pitch change and variants of the modulation spectrum. Using a corpus of conversational speech manually annotated for dimensions of stance-taking, we demonstrate that these measures of pitch and energy dynamics can help to characterize and distinguish among stance-related behaviors in speech.
Cite as: Levow, G.-A., Wright, R.A. (2017) Exploring Dynamic Measures of Stance in Spoken Interaction. Proc. Interspeech 2017, 1452-1456, doi: 10.21437/Interspeech.2017-1706
@inproceedings{levow17_interspeech,
author={Gina-Anne Levow and Richard A. Wright},
title={{Exploring Dynamic Measures of Stance in Spoken Interaction}},
year=2017,
booktitle={Proc. Interspeech 2017},
pages={1452--1456},
doi={10.21437/Interspeech.2017-1706}
}