Advances in IT are making possible new tools for human-human communication. Integration of speech, vision and dialog offers the possibility of a new class of tools to aid communication between people from different cultures using different languages.
The project will address the problem of integrating multiple communications modalities blending the physical and virtual worlds to provide support for multicultural communication and problem solving. The major challenges will be automatic perception of human action and understanding of human free dialog between people from different cultures. The consortium will construct an information butler, which demonstrates context awareness in a problem-solving scenario using computer vision, speech and dialog modelling.
The result will be a class of integrated tools for computer enhanced human-human communication. A public demonstration is planned for the Barcelona Cultural Fair 2004.
FAME's goal is to construct an intelligent agent to facilitate communication among people at separate locations from different cultures who collaborate on a common problem.
It will provide three services:
1) facilitate human-human communication through multimodal interaction including vision, speech and object manipulation;
2) provide appropriate information relevant to the context;
3) make possible production/manipulation of data blending electronic and physical representation.
It will serve as a transparent information butler in multicultural communication.
It will not intervene in conversations, but remain in the background to provide appropriate support.
A public demonstration is planned.
FAME is composed of 11 work packages, including management (WP1) and dissemination (WP11).
Three demonstrators of increasing complexity will be designed and the interfaces between the components of the system will be specified (WP2).
These demonstrators will concern scenarios for seminar style discussion, round-table group problem solving, and collaborative problem solving involving remote groups. Each scenario will include assistance for communication between people from different cultures using different languages.
A test-bed will be constructed to verify compliance with the specifications and to provide an environment for evaluating the performance and usability of software components (WP3).
At the end of the project, a robust demonstrator will be exhibited at the Barcelona Cultural Fair (WP2).
Constructing a facilitating agent for multicultural communication requires technological innovations in several areas:
1) Perceptual awareness of human activity using computer vision and speech understanding to control the acquisition and presentation of video (WP4) and audio (WP5) for communication;
2) Active control of acquisition to allow users to speak and move freely about the room unencumbered by microphones or wires;
3) Conversational speech recognition to recognise unplanned speech acts in order to feed the conversation context model with the meaning of utterances (WP6);
4) Dialog modelling (WP10) to maintain a task model that integrates information from perceptual awareness in order to inform the information retrieval process;
5) An information retrieval process to provide information relevant to the current communication and problem-solving context (WP7);
6) Natural interaction, blending physical and electronic representations of information, to enable people to speak, write and manipulate physical objects as part of the problem solving process (WP8);
7) Automatic speech translation to enable communication across cultural and language barriers (WP9).
Month 06: Interfaces designed and specified, test-bed available, evaluation criteria defined;
Month 12: First versions of all software components available;
Month 24: Optimised versions of all components are evaluated;
Month 30: Prototype Systems for each showcase are integrated and running;
Month 35: Demonstration at "Forum of Cultures", Barcelona 2004, 3 showcases demonstrate functionality: Seminar, Round Table, Distributed Workspaces.
Funding SchemeCSC - Cost-sharing contracts
75794 Paris Cedex 16
78153 Le Chesnay
38031 Grenoble Cedex 1
38400 Saint Martin D'heres
38400 Saint Martin D'heres