OK. Here's a long-promised report on the current status of speech-to-text according to my experience. Since I run Windows 7 on VMware on my Mac for a few specific tasks, I had to choose between Dragon Naturally Speaking v12 for Windows and Dragon Dictate 3 for Mac. I chose DNSv12 for Windows for two reasons: (1) Net cost to me to try the software and (2) DNSv12 for Windows is Nuance's flagship offering that is presumably their best foot forward.
The cost issue involved the fact that I already have two Sony ICD-MX20 digital recorders that have worked beautifully for my purposes for a long time AND they have the highest rating for accuracy among recorders that Nuance has tested. The digital voice files have a unique Sony format .msv (memory stick voice). They can be converted to MP3 files using Sony's software, but Dragon Dictate 3 Mac can't use them. It supports MP4 (plus other file formats) and not .msv or MP3.
Dragon Dictate 3 is a new offering that Nuance says is in the ballpark for accuracy with DNSv12, but I don't feel that it has had enough time on the market to establish a credible reputation.
My impression of DNSv12 after using it for one day only to transcribe speeches that were pre-recorded on the Sony units is that it is incredibly accurate for my purposes. Since the speeches are pre-recorded in front of other people, I'm not about to be saying "comma," "period," or "new paragraph." What I get in rough draft form is an accurate capturing of the words of the speech that need formatting and some word correction, but, IMHO it is an amazing result given the inherent variability in the way we speak. This is a huge step forward compared to the earlier version of DNS that I purchased and abandoned several years ago. Compared to the way I have been transcribing the speeches in the past, it is a huge improvement. Setting DNS v12 for maximum accuracy but slowest transcription rate produces a rough draft of two full single-spaced pages in 9 minutes. It takes me longer than that to transfer the .msv file to a flash drive and deliver it to the transcriptionist. It means that I have cut my total production time from a few days to a one-hour turnaround by the time I finish proofing and printing the final copy. I may decide to get a quality microphone and do some dictating, but I'm advised that the quality of the microphone is as important as the quality of the recorder, but, from reviews I have seen, a quality mike with a USB connection is a good way to go. So, at this point, I can't comment on the usefulness of office dictation of emails, memos and longer documents. Nuance's web site has some promotional videos showing these things in action, but I noticed that the persons who do the demonstrations have unusually clear diction. I doubt that I could match their accuracy in dictation. I hope that this is helpful to someone.
Edit: The report to this point appears like an ad for Nuance. There is a downside. It is that they hit you with a vocabulary that they have constructed to describe the various functions and procedures of DNS. It is something of a tech-speak snow job that isn't necessary to get a person doing a garden variety transcription either with a recorder or a microphone. I think they would get more enthusiastic buyers if they first introduced a person to a step by step procedure for getting simple results. Then, they could bring in the bells and whistles. It reminds me of the complexity of Windows as it added functions and grew and grew. For the past few months, I have been using a new iPhone 4S. Most of its functionality is presented in an intuitive way that enables the user to eventually figure most things out with very little help. That device stands in contrast to the complexity of the Nuance introduction to DNS.
Second Edit: In all fairness, they do have a help tab that lays out various topics, but my complaint is that DNS needs an intuitive, simple introduction up front to help the new user get the fundamental concepts and feel success before exposing the person to all the neat stuff with the fancy names.
Last edited by JoBoy; 10/24/12 09:40 PM.
