Prosodically Enriched Text Annotation for High Quality Speech Synthesis