• Automatic translation. Voice to voice and text to text translation for limited domains: Finite State Stochastic Transducers are used as the basis of the automatic translation systems. These models can automatically learn from real translation examples. Applications: translation of technical reports, hotel services, etc.
• Automatic recognition and understanding of speech: The audio signal is decoded into strings of words or semantic units. Finite state atutomata or grammars are used as the basis of those systems. These grammars can automatically be learned from examples of real pairs of signal–text. Applications: telephonic exchange services, control of devices through voice, informative petitions, etc.

Back to R&D&I lines