Speech expressiveness: Modeling and implementing the expressive impact of typographic and punctuation marks for textual inputs

The paper presents some validation results concerning the treatment of typographical and/or punctuation information in the perspective of an expressive speech synthesis for text inputs. In the first part, we outline the expressiveness model and the conceptual categories our work leans on and receives some legitimacy. This model incorporates a point of view of typographic and punctuation information and predicts their impact both in local and global expressive effects. In the second part, we give some issues concerning the developmental framework in which we work, specifically the TTS system of France Telecom. Our aim is to improve its expressive performances for text inputs. In the third part, we pinpoint the approach we adopt and put some light on the formal aspects we deal with in our treatments. In the fourth part, we sketch the user interface we developed for such an objective. We also try to give an idea concerning the obtained expressive improvements. In the final part (V and VI), we discuss validation issues and some perspectives concerning the most general expressive forms of the model and illustrate some of our first attempts to track them in our developmental framework.