What could be more natural than using our actions and words to be understood? Many groundbreaking scientific advances already enable us to contemplate the possibility of being able to interact naturally with machines. However, the limited standard of these "natural interfaces" currently restricts their practical applications. Notably, in terms of speech, processing data with regard to expressiveness according to context would bring potential new applications into play.
The Expression team's research project focuses on the study of data provided by people (in particular, linguistics) expressed in different ways: through actions, our movements, speech and written text. This information/data is made up of intrinsically complex features incorporating multidimensional and sequential elements. Additionally, these modes of expression are characteristically diverse. Some of them are symbolic (for example: words, phonemes, etc.) whereas others are numerical (positions, angles, sound samples, for example).
Within this complex context, in a broad sense linguistic data is made up of underlying structures from which the meanings emerge. The tone and expressiveness of the communication action can be made up from these elements. Translating this expressiveness and, more generally speaking, the variation in styles, into multidimensional time sequences, sequential data or linguistic structures represents the Expression team's main challenge. Working towards this objective involves assessing the feasible ways the variable elements of human communication and expressiveness data can be transcribed, especially with the aim of identifying constant patterns and improving characterization or data retrieval of expression categories.
The Expression team's work specifically focuses on the following areas of research:
- storage, processing and recognition of expressive gesture,
- expressive speech synthesis,
- expressiveness of text and languages.
The team is based at three locations: Lannion, Lorient and Vannes. In Lannion, the team primarily focuses on activities relating to expressive speech synthesis. Areas of particular interest include leisure (reading audio books, playing video games); education (mainly language learning); handicaps (speech impediments, teaching support).
- ROOTS : flexible multi-level word and text data structure transcription technology
- Expressive speech synthesis engine
- Recording studio