Once upon a time... GV-LEx
The GV-LEx project is at the crossroads of two fields of study and two recurring needs in our society:
- On one hand, we currently observe the growth of robots, humanoid or not, designed to have a social role (children's education and care, entertainment, etc.) or an assistance role (caring for the elderly or the disabled, etc). These robots are mobile and are becoming more and more adaptive. They also can begin interacting with their environment although the expressiveness required for an extended interaction, especially with childen, is still absent.
- On the other hand, the speech synthesis is more and more used in the context of information access (reading mails by phone, etc) or assistance (reading for the visually-impaired). The speech synthesis tools are efficient enough to achieve many tasks, however the expressiveness is not satisfying enough for a pleasant and persuasive reading experience.
The principal idea of GV-LEx is to make a robot able to read in a pleasant manner, with expressive capabilities based on both gesture and voice. Such a robot would have an important pedagogic value with both children and adults.
Three project objectives
- Develop a computer-based text comprehension tool (i.e. the meaning and structure) which allows for the interpretation of the text for reading.
- Synthesize an expressive voice.
- Generate expressive gestures that will be synchronized with speech synthesis.
Breaking down the barriers
- Identify in the text what can and what should generate expressiveness.
- Model the means used by human story tellers to transmit this expressiveness in their voice and gestures.
- Develop speech corpora to synthesize an expressive voice.
- Develop a library of expressive gestures usable by the robot and the avatar.
- Synchronize gesture and voice.
- Develop a tool for manual modification of computer-based interpretation.
Scenario
- The robot tells a story to 7 or 8 year-old children for a maximum of 2 to 5 minutes. No interaction with the robot is planned. At the end, the text is pre-processed off-line with a software specially developed for GV-LEx.