Voice Module Work Plan

We plan to have a "wizard of oz" deployment in Bilaspur which is
manned by a human operator, for a while.  The human operator will do
the job that the computer is eventually supposed to do.  We will
record the audio from all transactions. We will also record the
operators responses, when he/she made mistakes etc. The audio
recordings will be used to train the speech recognition system and the
NLU. The operator's responses will be used to learn about how the
system must respond.

We will use these data to develop the various components of our
system, based on previously developed tools.