ISCA Archive SpeechProsody 2014
ISCA Archive SpeechProsody 2014

Towards Automatic Extraction of Prosodic Patterns for Speech Synthesis

Mónica Domínguez, Mireia Farrús, Alicia Burga, Leo Wanner

This paper deals with the adaptation of AuToBI annotation for speech synthesis purposes. AuToBI is a tool that automatically detects and classifies the standard ToBI labels for American English. AuToBI annotation is performed word-by-word. However, a labeling of intonation patterns at the intonational phrase level is essential for the detection of the correlation between theme/rheme (thematicity) and prosody and also much more appropriate for speech synthesis applications that use various layers of linguistic annotation (syntax, semantic, information, and prosody structures), such that if used in speech synthesis applications, AuToBI’s output would require a post-processing stage of the extracted labels. We present a procedure that includes an initial AuToBI annotation and the adaptation of the AuToBI output to a phrase-based annotation, following a set of determined rules. A further analysis of the correspondence between prosodic patterns and themacity structures is used to validate the results.