Knowledge about intonation - the ups and downs of the voice in speech - is increasing rapidly. The main reason is theoretical: researchers are now paying attention to targets (peaks and valleys) rather than pitch movements (rises and falls). There is also a practical motivation: successful speech synthesis now mostly uses target-based models to create natural-sounding intonation (instead of robot monotone).
he target theory is now being applied to numerous languages; our project deals with Modern Greek. Our focus is on theoretical issues concerning targets. By studying the temporal alignment of targets (known to be a significant phonetic variable), we hope to distinguish primary (relatively fixed) targets from secondary (more variable) ones. We also hope to establish a distinction between primary and secondary accents (combinations of targets). The empirical research will consist of production experiments (acoustic measurements of carefully selected spoken utterances), followed up by perceptual experiments (where Greek speakers judge utterances with experimentally modified intonation).
The project also aims to produce a full description of Greek intonation that can be used in speech synthesis. This goal is relevant to the increased interest in multi-language speech technology within Europe.