The goal of the POLYGLOT project was to demonstrate, through research, the feasibility of the multi-lingual voice interface (input/output) for a number of commercially promising applications. The objective of the project was to integrate phonetic, lexical, and syntactic knowledge common to text-to-speech and speech-to-text conversion, providing greater generality, lower cost, and easier extensions. POLYGLOT also aimed to provide the results of the research and applied technology on a PC.
The goal of the project was to demonstrate the feasibility of multilanguage voice input/output for a number of commercially promising applications. The objective was to integrate phonetic, lexical, and syntactic knowledge common to text to speech and speech to text conversion, providing greater generality, lower cost, and easier extensions.
A set of technologies and working prototypes was planned:
a very large vocabulary, isolated words and speaker adaptive speech recogntion on a personal computer (PC) for speech to text conversion in 6 European languages (continuous speech recognition will be addressed as well, but only to the stage of producing a strategy for implementation on a PC);
high quality, language sensitive, unrestricted vocabulary with text to speech conversion in 6 European languages;
tools for easy extension of both the recognition and synthesis technologies to additional languages;
application development tools for recognition and synthesis;
the speech to text system (now available under the popular Windows environment);
development of demonstrators for languages teaching, speech to text medical report preparation with voice feedback, remote access to an electronic mailbox, access to a mail directory, a generalized Windows interface application, and a telephone directory access application.
In addition, the project has produced speech databases for speech to text and text to speech for 6 European languages. These are being consolidated and will be available for further development work.
6500 HD Nijmegen