Сегментація на акустичні склади для компілятивного синтезу англійського мовлення

Abstract

This paper describes the approach to syllabification currently used in the Laboratory of Computational Linguistics (Kyiv National Linguistic University) for English text-to-speech synthesis. Syllable is a subcomponent of the word model whereas quasi-syllable is a subcomponent of the hierarchic speech model of phonetic word, rhythmical group, and syntagma. The real structure and syntactical position of quasi-syllables enables the TTS-system to retain the naturalness of concatenative speech signals.