Preliminary results on the performance of a system for the automatic recognition of continuous speech.