Synthesized Spoken Names: Biases Impacting Perception

Lucas Kessler, Cecilia Ovesdotter Alm, Reynold Bailey


Utilizing a existing neural text-to-speech synthesis architecture to generate person names and comparing them to reference names read aloud in a formal context, we explore how bias resulting from training data impacts the synthesis of person names, focusing on frequency and origin of names. Long-term, we aim to apply voice conversion of person names to aid the effective reading aloud of such names in celebratory ceremonies.


Cite as: Kessler, L., Alm, C.O., Bailey, R. (2019) Synthesized Spoken Names: Biases Impacting Perception. Proc. Interspeech 2019, 3689-3690.


@inproceedings{Kessler2019,
  author={Lucas Kessler and Cecilia Ovesdotter Alm and Reynold Bailey},
  title={{Synthesized Spoken Names: Biases Impacting Perception}},
  year=2019,
  booktitle={Proc. Interspeech 2019},
  pages={3689--3690}
}