The TALK project focuses on the development of new technologies for adaptive multimodal and multilingual human-computer dialogue systems. We are working towards the long-term vision of users interacting naturally with devices and services, in the home or car, using speech, graphics, or a combination of the two.
As a starting point, we adopt a successful approach to dialogue management called the "Information State Update'' (ISU) approach, which was developed in TALK's predecessor projects, and is now used in many projects and applications worldwide. The ISU approach supports a much richer representation of dialogue context than finite state machines or Voice XML - the approaches most commonly used to build industrial dialogue systems. In TALK we are extending the ISU approach to support our two central objectives: developing reusable systems that are adaptable to different languages, graphical modalities, and application domains; and improving the usability of dialogue systems by employing techniques to learn optimal system behaviour from human dialogue data.
In developing new technologies, the academic partners are working closely with industrial partners (Linguamatics Ltd., BMW F&T GmbH, Robert-Bosch GmbH) for insights regarding the shape that such technologies should take, so that TALK results can be directly integrated into the industrial partners' research and development.
To demonstrate these new technologies, we are building showcases for incar and in-home information and control - areas in which our industrial partners are actively involved.