[ninjin]

As a somewhat established researcher in the field, I second Jurasky and Martin. It is peerless and what I recommend to anyone joining my team if they think their background NLP knowledge is a bit on the weak side.