Can you clarify how you expect to use the recorder and its recordings? I ask because using DNS to transcribe the recordings you describe from multiple sources is not likely to be successful, so there's no particular reason to limit yourself to just digital recorders. But if you intend to transcribe your own voice recordings, that's a different, and more feasible, goal.

Bruce

Anthony,

Back before I left ScanSoft in January of 2002, I was responsible for establishing protocols for digital voice recorders. One of the things that I reported to QA and development was that there were a lot of things that you could do with a digital voice recorder that were not publicly discussed, but for good reason. L&H/ScanSoft/Nuance technical support obviously does not want to endorse some of these procedures and capabilities.

However, as regards what you want to use the recorder for, in the words of us New Yorkers, fagetaboutit !!!:jawdrop:

Unfortunately, although there are ways of using DNS to transcribe multiple speakers, there are basic things that go without saying.

1. Lectures, conversational speech, non-structured speech, conferences, meetings, etc. all involve the type of speech that is not well suited to recording using a DVR (Digital Voice Recorder). Crosstalk and over talking, as well as the type of conversational speech that is involved in these types of environments are not likely to produce anything better than 5% to 10% accuracy. Speech recognition is simply not designed to handle these types of speech environments, even though the latest versions of DNS (9.5-10.0) do come with speaker independent Acoustic Model (Skip initial training for this user – DNS 9.5 and selecting "None" when creating a new user in DNS 10.0). These speaker independent Acoustic Model's still require that the speech being transcribed is structured and not conversational. You can test this very easily by simply talking to DNS the same way that you would talk to somebody in a normal standard conversation with all of the various slang and colloquialisms. You'll find out very quickly that the results end up being mostly garbage. So, using a digital voice recorder for this purpose is not going to net you results worth spending the money for a good DVR. Under these conditions, you're better off doing what is called Echo dictating. Echo dictating is basically listening to a recording and then dictating what you hear using your own user profile. The amount of time and effort required to correct inaccurate recordings of lectures and the like will take you twice as long as echo dictating would take. Bottom line: It won't work. It is virtually impossible to structure these types of environments in a manner that would be conducive to transcribing via DNS and a DVR.

2. Recording any type of external source requires that the clarity be high, the speech be structured, and the volume be optimal without a lot of background noise, cross talking, multiple speakers speaking at the same time, or, what I call, the canyon effect (i.e., your recordings sounding like you're talking in a canyon). But all of these factors together and you get very poor results.

3. When using a digital voice recorder for recording any external source for transcription, you cannot use the automatic voice activation feature because this will cause dropouts (i.e., sections of your dictation that will be missing parts of what is being recorded). In addition, you cannot use the conference setting because using this type of setting on a digital voice recorder amplifies the input to the microphone. The problem here is that it also amplifies everything else, including fans, people talking around you, people talking across the room, etc., etc. ad infinitum. This is anathema to transcribing any recordings made under this condition.

4. On the other hand, if you're going to use a DVR professionally to record your own personal dictation for transcription, then the expense of purchasing a good digital voice recorder may be justifiable. Otherwise, the only recommendation that I can give you is to use a reasonably good recorder that gives you clear reproduction of that which you are recording so that you can use the echo dictation technique. It's not worth spending $500-$800 on a digital voice recorder under the conditions that you specify.

Chuck Runquist
Former Dragon NaturallySpeaking SDK & Senior Technical Solutions PM for DNS for Lernout & Hauspie (L&H)

"Some have taken authority as the truth, rather than truth as the authority." -- Gerald Massey