View article

tion of this technology is the rate of errors and lack of graceful error handling [8]. Benchmark error rates reported for speech recognition systems are still too high to support many applications [4]; the amount of time users spend resolving errors can be substantial and frustrating.