Natives’ and non-natives’ pronunciation of Dutch, and it’s influence on automated speech recognition
Summary
The Autonomata and Autonomata TOO research programs have produced large annotated speech corpora mainly targeted for Personal Navigation De- vices. In this paper these corpora are analyzed. Systematic relationships between subjects’ mother language and the deviation of the realized pronun- ciations to the standard pronunciation are identified, and the effects this has on the accuracy of the Automated Speech Recognition system that was part of the Autonomata projects.