SpeechRec contains models required by the isr
Currently they are derived from acoustic models trained with the Verbmodbil (de)
and the Wall Street Journal Corpus (en)

Structure:
./de -- everything based upon german models
./en -- everything based upon english models

* Each subdirectory within these contains exactly one set of files required to
  build models for the isr.
* build.sh documents the call to build/rebuild each model
* MakeVars.models contains set specific settings (files with phonetic
  transcription, grammar name, base model, ...)

***** Howto create a new set *****
1. Use copyset.sh to copy one of the existing sets to a new one
   !!! Do not use cp without additional options !!!
   !!!      Symlinks need to be preserved       !!!
2. Delete all unnecessary stuff from the new set which has not already been
   removed
3. Modify MakeVars.models and Makefile accordingly