Now we have all the components, it will be useful to see them all working together.
The generative view
Yes, again! Really, this is the best way to see how everything integrates elegantly, just by multiplying probabilities.
Compiling the recognition network
The idea of compiling al the models together is very natural, if we are taking the generative view.
Training
The Baum-Welch algorithm can be applied to data comprising whole utterances transcribed with words, without needing any manual within-utterance alignment.
