Assigning phrase breaks using CARTs for Basque TTS

Eva Navas, Inmaculada Hernáez, Nerea Ezeiza

This paper presents a prosodic phrasing method for the Basque language, to improve naturalness in text to speech synthesis. Binary classification trees are trained with morphological and syntactic information to predict locations of breaks. Overall score achieved by the prediction tree is 92.53%, which compares positively with the results published for other languages.

Cite as: Navas, E., Hernáez, I., Ezeiza, N. (2002) Assigning phrase breaks using CARTs for Basque TTS. Proc. Speech Prosody 2002, 527-530

@inproceedings{navas02_speechprosody,
  author={Eva Navas and Inmaculada Hernáez and Nerea Ezeiza},
  title={{Assigning phrase breaks using CARTs for Basque TTS}},
  year=2002,
  booktitle={Proc. Speech Prosody 2002},
  pages={527--530}
}