Speech synthesis technology:
'Parameterized' - relatively small and fast (low
quality)
Concatenative - more complex and resource intensive
(higher quality)
Speech recognition technology:
'Speaker-independent' systems (no training, low
scope and reliability)
Training-set based (tuned to a particular user)
Discrete word vs continuous utterance recognition
Variety of techniques used:
Fixed-phrase "command and control"
Rule-based grammar systems
Statistical models
Neural-network assistance