Context-dependent classes in a hybrid recurrent network-HMM speech recognition systemThe 1995 ABBOT LVCSR System for Multiple Unknown Microphones
