An enhanced pitch modeling supporting a Greek text to speech system

This paper tries to describe as accurately as possible an enhanced procedure for predicting the appropriate prosodic structure of speech and apply it to a Greek Text to Speech system. The main focus will be a particular linear stylization of the fundamental frequency function F0 contour (pitch), by using the most important syntactic, grammatical and lexical features of the Greek language. Finally, we present an evaluation on the results of plugging the pitch model into the system.