A Japanese text-to-speech system based on multi-form units with consideration of frequency distribution in Japanese.