Perceptual Adaptation to Device and Human Voices: Learning and Generalization of a Phonetic Shift Across Real and Voice-AI Talkers

Bruno Ferenc Segedin, Michelle Cohn, Georgia Zellou


Voice-activated artificially-intelligent digital devices are a new type of interlocutor. Like for human talkers, they have idiosyncratic speech patterns that require listeners to perceptually adapt to during language comprehension. One question is how perceptual adaptation to a novel accent in speech produced by a digital device voice compares to adaptation to human voices. Furthermore, adaptation to one talker can generalize to novel voices. Hence, we also tested whether perceptual adaptation to accented device voices generalizes to novel human voices, and vice versa. In this study, listeners were first exposed to words with a shifted phoneme realization in either a device or human voice. Later, participants were tested on whether they shifted their identification of words in the shifted talker. Additionally, we tested whether listeners applied the shift to novel device and human voices not heard in exposure. Results reveal talker-specific learning for both device and human voices. Yet, the size of the shift was larger for the device voices. Furthermore, listeners exposed to the shift in device voices showed generalization to novel human voices, and vice versa. These patterns of adaptation and generalization for device and human talkers have implications for models of speech perception models and human-computer interaction.


 DOI: 10.21437/Interspeech.2019-1433

Cite as: Segedin, B.F., Cohn, M., Zellou, G. (2019) Perceptual Adaptation to Device and Human Voices: Learning and Generalization of a Phonetic Shift Across Real and Voice-AI Talkers. Proc. Interspeech 2019, 2310-2314, DOI: 10.21437/Interspeech.2019-1433.


@inproceedings{Segedin2019,
  author={Bruno Ferenc Segedin and Michelle Cohn and Georgia Zellou},
  title={{Perceptual Adaptation to Device and Human Voices: Learning and Generalization of a Phonetic Shift Across Real and Voice-AI Talkers}},
  year=2019,
  booktitle={Proc. Interspeech 2019},
  pages={2310--2314},
  doi={10.21437/Interspeech.2019-1433},
  url={http://dx.doi.org/10.21437/Interspeech.2019-1433}
}