Model training and evaluation scripts

Overview

This is a collection of scripts for training and running speech recognition and (one day) synthesis. It is largely based on HTS rather than HTK.

The scripts themselves are written mainly in zsh, which is like sh or bash, but it handles arrays like csh. Basically, if you know bash then just pretend it's bash. The place where this will come undone is arrays: In bash the first element is ${array[0]}, in zsh it is $array[1]. bash's ${array[*]} is just $array in zsh. zsh is nicer :-)

These scripts will run under SGE simply by setting

export USE_GE=1

In this case, the sequence of working scripts called at top level form a graph of SGE jobs. Each working script is held until the previous one completes.

Documentation

For more documentation, please see the iss wiki on GitHub.

There is no mailing list (yet), but feel free to use the juicer list: To subscribe, send a message to mailto:[email protected] with subscribe in the body.

Phil Garner, Idiap, October 2010