Key points are not available for this paper at this time.
By making the advances necessary to implement next-generation speech recognition applications, researchers could develop systems within a decade that match human performance levels.
Padmanabhan et al. (Tue,) studied this question.