2

Dragon NaturallySpeaking's editor Nuance claims that Dragon 12 delivers up to a 20% improvement in out-of-box accuracy compared to Dragon 11. Nuance made similar claims in the past with previous versions. Are there any rigorous studies to verify this commercial claim?

Franck Dernoncourt
  • 7,224
  • 4
  • 33
  • 76
  • 3
    I tend to that there will be peer-reviewed studies of the efficacy of the commercial product, especially one that is in constant development. Anecdotally, I can say that the quality of dictation has improved between Dragon generations, but the software still makes frequent errors, some of which are easy to fix, others, not so much. <- This comment transcribed by Dragon dictate without correction. – Larry OBrien Dec 24 '13 at 01:00
  • [This informal test](http://blog.classroomteacher.ca/750/how-accurate-is-dragon-naturallyspeaking-11-using-the-rainbow-passage-to-measure-accuracy/) found, at its worst, Dragon 11 was 87.6% accurate out-of-the-box.. What does it mean to offer a 20% improvement on that? – Oddthinking Dec 24 '13 at 04:53
  • 1
    @Oddthinking Some people mean a 20% reduction in error rate, which in this case, would be reducing an error rate of 12.4% to 9.9%. –  Dec 24 '13 at 07:21
  • 1
    @Articuno exactly, use whichever number that looks larger and more impressive for marketing; no-one checks them anyway ;) – ratchet freak Dec 24 '13 at 13:12
  • @Oddthinking I don’t think that’s a meaningful number. 87% accuracy is abysmal and altogether useless for practical application. You need accuracy upwards of 98% (or thereabouts) for it to be a useful product. So the out-of-the-box accuracy doesn’t seem to be an informative number. – Konrad Rudolph Dec 24 '13 at 18:22

0 Answers0