Project team CORDIAL

Scientific Context

• Man machine communication
• Automatic learning
• Speech technology: speech recognition and synthesis

Activity report

Research activities

Activities of the CORDIAL project are divided into two main complementary areas with the aim of introducing user-friendly interactive spoken systems. The first area includes fundamental studies, the completion of which is considered a prerequisite for automatic system component design. The second area focuses on the practical aspects of the implementation, integration and assessment of spoken interactive systems.

Fundamental research includes modelling different aspects of the man-machine communication: dialogue models (structural, purposeful), reference processing, communication error processing and mediation dialogue. They include the automatic learning of syntactic, prosodic, and dialogue structures: which aim to make the implementation of new systems easier. Finally, prosody studies are being developed in an educational perspective and to improve speech synthesis.

Application projects have enabled us to develop several prototypes for interactive spoken systems (multi-modal database queries, educational packages). The difficult problem of evaluating dialogue systems is also under study.

Prototypes

• Touch-sensitive Géoral: multi-modal interrogation system based on speech, gestures and graphics, for geographical and tourist information databases.
• Ordictée: spelling tutor based on dictation with speech synthesis