A new intonation model for inflected verbs in Turkish text-to-speech synthesis

In this study, we aim to build a rule-based intonation model for Turkish text-to-speech synthesis. A proper intonation model is required for a natural synthesized speech. Since verbs in Turkish are very important in the sentence, and their inflection rules are well known, the derived intonation rules are based on verbal inflections. Stress patterns of affirmative, negative and interrogative forms of verbs are investigated in terms of combinations of duration, pitch frequency and energy modifications. In the end, very much related with the quality of the synthesis, it is found that, for positive and negative affirmative sentences, no parameter differs much; nevertheless, for positive and negative interrogative sentences, pitch frequency and energy are seen to be the most important and effective parameters.