A Chinese text-to-speech system based on part-of-speech analysis, prosodic modeling and non-uniform units.