Ekaterina Saenko, Nick Campbell
Recording New Databases For
CHATR
Abstract:Producing new speech databases for CHATR, a multi-speaker, multi-
lingual speech synthesis system, is an important task which in the
past had been accomplished by having a speaker read an arbitrary text.
This paper discusses ways to facilitate the recording process, in
particular, the selection of the input text based on a set of phonetic
units and a phonotactic model of the target language, and outlines
the design of an interactive recording and labeling tool.