WORD IDENTIFICATION METHOD FOR JAPANESE TEXT-TO-SPEECH CONVERSION SYSTEM.

A word identification method is proposed for Japanese text-to-speech conversion. The number of identification trials is reduced by restricting the identification to short intervals using a dynamic programming technique. Plausibility evaluation that reflects several Japanese text writing characteristics is introduced. Results show the usefulness of this identification method as well as the necessity of dictionary enrichment and of advanced intensive analyses. These advanced analyses are not only required in word identification itself but in the further text analyses needed for the generation of phonetic and prosodic information.