The Calista recognizer that was used was derived from Sphinx-3.
•
The SpeechBot indexing workflow involved a farm of Windows workstations that retrieved the streaming content; and a Linuxcluster running speech recognition to transcribe the spoken audio.