Spoken Language Translation (SLT) Demonstration

Welcome to SRI's Speech Technology and Research Laboratory (STAR)

This demonstration lets you ask questions in English about airline flights between selected cities, flight times, fares, etc. The system will try to translate what you say into French.

To do this, the system integrates many different technologies: speech recognition, intelligent agents, natural language understanding, machine translation and speech synthesis.

If you are unable to reach the following phone number, check out the VIDEO

How to Start the Demonstration

You must have a browser that supports Java, and access to a telephone with which you can call +1 (650) 328 5324 in the United States.

About the technology

Speech Recognition

The demonstration is powered by the Nuance Speech Recognition Engine, developed by Nuance and SRI STAR Lab. Custom applications can easily be developed using the Nuance speech recognition toolkit. For more information on SRI's multilingual, speaker independent, speech recognition technology please contact Leo Neumeyer.

Machine Translation

The language analysis and machine translation module is being developed by SRI Cambridge (UK). Please contact Manny Rayner. It is written in prolog and can handle Swedish as well as English and French, as shown in this demonstration.

Speech Synthesizer

The demo uses the CNetVOX French speech synthesizer by ELAN Informatique.

Agent Architecture

All these components are realized as software agents that exist and function independently of each other, and possibly on different computers on the Internet. This architecture is called the Open Agent Architecture and is developed by SRI's Artificial Intelligence Center. In this demonstration, the agents are written in C, Java and Prolog.

Acknowledgments

The work was funded by SRI International and Telia Research AB. The Web demonstration was developed by Erland Lewin while visiting SRI STAR Lab from the Department of Speech, Music and Hearing at the Royal Institute of Technlogy. We greatly appreciate the help given by Luc Julia, Adam Cheyer, Mehdi Charrafedine and David Carter.