How can we build recognizers for quickly? How can we learn automatically what sound units there are, how they are pronounced, what words there are - and build a recognizer, or at least a system that can detect a few key terms?

More to come...