Sixth ISCA Workshop on Speech Synthesis

Bonn, Germany
August 22-24, 2007

Building a Better Indian English Voice Using "More Data"

Rohit Kumar, Rashmi Gangadharaiah, Sharath Rao, Kishore Prahallad, Carolyn P. Rosť, Alan W. Black

Language Technologies Institute, Carnegie Mellon University, Pittsburgh, PA, USA

We report our experiments towards improving an existing publicly available Indian English voice using additional data. The additional data was used to create new duration and pronunciation models as well as to convert the existing voice to create a more Indian sounding voice. Two experiments along the above lines are reported. In the first experiment, we found that changing the pronunciation models has the potential to improve an existing Indian English voice. We conducted a second experiment to validate this finding. The second experiment shows the potential value in carefully investigating the separate effects of the different components of a pronunciation model in order to understand their unique contributions to improving an Indian English voice.

Full Paper   Poster (pdf)

Bibliographic reference.  Kumar, Rohit / Gangadharaiah, Rashmi / Rao, Sharath / Prahallad, Kishore / Rosť, Carolyn P. / Black, Alan W. (2007): "Building a better Indian English voice using "more data"", In SSW6-2007, 90-94.