Trys to mimic human-human-audio communication, thus takes a command based approach.
Assumes English, German or French language, but a change to another language should be trivial.
Minimum device capabilites:
Audio (wave) I/O
Reliable speech generation from text
Reliable speech recognition based on an application-provided highly dynamic vocabulary with several thousand entries
Speech recognition of freetext based on an huge vocabulary
The user speaks commands in sentences. Output is also generated as speech.