A 14% word error rate is currently considered the world's best Automatic Speech Recognition (ASR) for this kind of speech, which measures the percentage of words that are incorrectly transcribed.