 |
ESCA Workshop on ProsodyLund, Sweden |
 |
|
ESCA Workshop on ProsodyLund, Sweden |
|
This article presents a complete algorithm for the generation of intonation (F0 contours) for the Greek Text-To-Speech system, based on a multi-layer label structure that is constructed for the phonemes representing the input text This structure consists of the phoneme's distinctive features, the position of the syllable that the phoneme belongs to, the prosodic label of the word that the phoneme belongs to, and the phoneme's prosodic context in the sentence. According to the contents of that structure, the algorithm assigns to each phoneme of the input sentence a target pitch level to be reached either at the beginning, or the middle, or the end of the phoneme. When all the phonemes have been assigned the appropriate F0 level, the overall pitch contour is constructed by linear interpolation between the successive F0 levels. Although the method proposed seems to be a rather abstract approach, it takes into consideration linguistic, phonotactics and metrical constraints of the input and not linguistic constraints alone. In addition, the method is especially suited for languages, such as Greek, which are inflectionally rich and have great freedom of word-order.
Bibliographic reference. Epitropakis, George / Yiourgalis, Nikos / Kokkinakis, George (1993): "High quality intonation algorithm for the greek TTS - system", In Prosody-1993, 70-73.