Automated Multilingual Interaction with Information and Services - AMITIES

AMITIES is an international EU-US collaboration project which aims to develop, showcase and deploy novel technologies for building empirically induced dialogue systems to support automated multilingual human-computer interaction with information and services.

The central thrust of the work is the synthesis of dialogue systems from large corpora of recorded human-human and human-computer conversations. The project aims to extend techniques that have been successful for creating robust parsers and speech recognisers to the area of dialogue systems by utilising probabilistic models and learning technologies. The AMITIES work is founded on the belief that such methodologies have matured to the point where they can be applied to efficient and reliable automated call center systems. Once such technologies are operative at the dialogue level, then they can be extended to new languages, applications, and user groups.

AMITIES aims to enable EU citizens to continuously access information and assistance from commercial companies via the telephone and the Internet using their native languages in a natural interactive dialogue.

The results of the project will be showcased at the commercial call centers operated by the consortium members across the EU and the US. The consortium will seek out other potential applications, such as health insurance services and public assistance applications. Data will be collected for evaluation and demonstration purposes, ensuring the future exploitability of the results.

OBJECTIVES

The project is developing technologies which will gradually replace traditional response systems with an automated natural dialogue system. Natural interactivity through a multilingual conversational interface will enable customers to retrieve information and obtain assistance quickly and efficiently:

• Customers calling an automated call centre will be greeted by an automated Call Answering and Routing Triage (CART) which conducts a natural language dialogue in the customers' language, to determine the reason they are calling.
  
  
• The dialogue can be spoken (traditional telephone), or multi-modal (internet/phone) including speech, type, and point-and-click input, and speech, text, graphics, and video output.
  
  
• CART will route the call to the appropriate back-end destination that includes human operators and Automated Service Counters (ASCs).
  
  
• ASCs conduct task-oriented dialogue with the customers and help provide the information they seek. They rely on domain knowledge to solve problems.
  
  
• ASCs support a variety of specific services, which are gradually added to the system. Tasks covered range from retrieving account balances, to credit dispute resolution, to loan application interviews, etc.

Project factsheet optimized for printing

Last updated: 22.11.2002

[  Projects Home  |  Alphabetical List  |  Programme List  |  Group List  |  Search Projects  ]