summery
Considering the steadily increasing dissemination of computer applications and their growing functional complexity, it is crucially important to avoid that usage and control of these systems and services gets more and more complicated. To this end the utilization of natural language as a means of operation is a logical consequence to facilitate their usage. Thus user-friendly, comfortable and safe vocal interfaces are needed that ensure natural interactivity. VICO – Virtual Intelligent Co-Driver – aims at the development of an intelligent conversational agent enabling ubiquitous natural interaction between humans and digital devices and services.
Focusing on the automotive environment, the project will provide a user-friendly natural language interface for in-car devices and innovative services, such as navigation, on-the-fly route planning, interactive hotel and restaurant reservation and tourist information.
To realize these functionalities the project will pursue the following technical objectives:
• development of VICO, a virtual intelligent co-driver enabling natural interaction between humans and digital devices and services in the car
• providing a user-friendly natural language interface for in-car devices and innovative services, such as navigation, on-the-fly route planning, interactive hotel reservation, etc.
• integration of:
◦ robust spontaneous speech recognition technology for adverse environments
◦ powerful and adaptive dialogue management strategies
◦ natural language understanding techniques for conversational interaction
◦ access to a dynamic information source and management of relevant data
◦ integration and evaluation of the developed technologies into an in-car demonstrator
Built in two stages of development a first prototype of the virtual co-driver will be available first, and at the end of the project a final demonstrator. These prototype demonstrator systems will integrate all approaches developed during the project, and will also be able of fulfilling all features presented below. User evaluations will be conducted in simulated as well as real driving situations all throughout the duration of the project in order to improve the design and usability of the emerging final system. Both activities are foreseen to be able to integrate user feedback into the evolution process as early as possible as well as to ensure acceptance by possible users of a virtual co-driver. Furthermore, the results of these evaluations are a valuable input for further development of future natural language conversational systems in adverse environments.
Project Facts
Duration: March, 1st 2001 – May, 31th 2004
Applications:
Languages:
• English
• German
• Italian
Contact Information
Dr. Tom Geutner
Robert Bosch GmbH
P.O. Box 10 60 50
D-70049 Stuttgart
Tel: (711) 811-6850
Fax: (711) 811-7584
Email: Bill.Geutner@de.bosch.com
objectives
Whereas the overall objective of VICO is the design and development of an intelligent agent providing a conversational speech interface to allow natural interaction between humans and electronic services, the focus is on solving the particular challenges in the automotive environment.
Special care is taken to create generic techniques and tools extendable to additional functionalities that might even go beyond project activities. Great importance is also attached to developing tools and techniques portable to other domains, like e.g. mobile agents within related applications. As such all of the following objectives are of key importance:
• allow natural language conversations through an intelligent agent
• create a user-friendly, comfortable, efficient and safe-to-use vocal interface
• develop robust speech recognition technology for adverse environments and unrestricted vocabular
• create a robust natural language understanding component to be easily adapted to other domains and application
• develop intelligent, flexible, and user- and situation-adaptive dialogue management strategies that are extendable within a specific domain and can also be ported to other domains
• offer multilingual conversational information and communication services within Europe
• set standards for natural language conversational human-machine interfaces
innovation
The goals of the project are the design, development and evaluation of:
• robust speech recognition methods for spontaneous speech in adverse environments (e.g. a car)
• adaptive dialogue management strategies and natural language understanding techniques for conversational interaction
To reach these goals significant improvements and extensions of existing echnology are necessary. Given the attentional load of a user driving a vehicle while talking to the virtual intelligent co-driver, the processing of natural speech requires high-performance speech and language processing techniques, and a flexible and adaptive dialogue management. Furthermore, the realization of a dynamic information source – the so-called CarNet – needs advanced data extraction and management techniques.
As a consequence, research in the following areas is of key importance:
• robust recognition of spontaneous speech in the automotive environment
• pronunciation rules for names in foreign languages
• robust natural language understanding
• integration of advanced dialogue techniques, optimised for handling of incomplete or ambiguous input
• generation of clarification dialogues
• representation of geographic, tourist and car-specific information for efficient access from the dialogue manager (CarNet)
• integration of all developed components into an in-car demonstrator
• user-level evaluation of intermediate laboratory prototype
• driver-level evaluation in real driving situations
partners
Robert Bosch GmbH – Germany
DaimlerChrysler AG – Germany
Istituto Trentino di Cultura – Italy
Phonetic Topographics N.V. – Belgium