I agree with everything Bruce says, and I used it in my practice since
1992 - and his experience mirrors mine exactly as he states.
I am worried, however, that in my haste, I may have missed John's
essential purpose in asking about this software. I re-read the post,
and now see he MAY want to use it to interpret other operators' speech
coming in on SSB, and not, say, perhaps, for use in dictating his
message for digital mode transmission, i.e., incoming vs outgoing
messages.
As for over the air signals with other voices... I predict it will be
very poor, indeed. While the new version 11 claims to be speaker
independent (at least nominally- they do say, as Bruce indicates, it
works better with even a short training session to tailor it to your
voice) but another voice, with all that static, and the inherent
distortion on SSB... well, that is not going to work well at all. In my
estimation, it will not work as well as computer CW translator programs.
Also, while it learns your speech patterns, high accuracy is dependent
on good speech habits, clear enunciation and completeness. For example,
if you slur your words together, like "thisanthat" it will not be as
accurate as if you spoke each work completely and made sure you
enunciate the "t" on the end of "that" and the "d" on the end of "and"
and so forth. I doubt SSB audio input would be sufficiently clear, and
well enunciated, to be used as a speech-to-text interpreter. In court
rooms they either record it and transcribe / re-dictate it later, or use
a "steno-mask" and re-dictate it live in the court room. This is
because the program does not, as Bruce indicates, process many different
voices at once.
I am surprised at how few writers are using this valuable tool - it is
invaluable for dictating quick notes, email, office correspondence, and
other stuff - especially where total accuracy is not always required, in
addition to formal work, such as contracts, deeds, pleadings, motions,
briefs, and other legal documents where accuracy is essential.
Just MY take. Happy days.
====================== James - K8JHR =======================
On 12/6/2010 12:21 PM, Bruce McLaughlin wrote:
I can second the recommendation for Dragon naturally speaking. I have both
the program for a PC and another version for the Macintosh. Both of them
work pretty much the same as far as recognition accuracy is concerned,
although, I think the interface for the PC is better. Version 11 which is
the current version for the PC is extremely good. I use it for legal
writing and for ham radio I have used it to work PSK 31 as well as RTTY by
dictating into various programs such as Mix W. In fact, I am using it to
write this message. Earlier versions of the program required a fair amount
of "training" in order to build up the so-called profile of your voice, that
is, how you pronounce words and familiar contexts you use. However, the
current version requires hardly any training and will, in fact, work very
well with no training at all. It "learns" how you speak as you use it. If
it makes a mistake on a word you correct it. It keeps track of those
corrections and when you close down the program it will add those
corrections to your profile so that it will not make that mistake again.
However, as someone has mentioned, it is still rather dependent on your
particular voice characteristics as well as a good clean audio input. I
don't know if it will work very well with over the air signals with static,
interference, and fading, not to mention various accents, etc. However, the
program is not very expensive and is undoubtedly worth a try. At the very
least, it should be an interesting addition to your software library and
could be quite helpful in ordinary writing. It comes with a very nice
headset mic which seems to work very well. I note that the program has made
absolutely no mistakes while I dictated this message. Considering what it
has to do I think that is amazing. I began using speech recognition with a
program put out by IBM called Via Voice about 10 years ago. It was very
expensive and very crude compared to NaturallySpeaking of today. Not only
is today's program much less expensive it is much, much better. They have
certainly made vast improvements in the technology during the past 10 years.
Good luck. Bruce-W8FU
-----Original Message-----
From: tentec-bounces@contesting.com [mailto:tentec-bounces@contesting.com]
On Behalf Of John Chance-Read
Sent: Monday, December 06, 2010 7:49 AM
To: tentec@contesting.com
Subject: [TenTec] Speech to Text
Being deaf, I am trying to locate a 'Speech to Text' software package.
All my attempts, despite careful choice of route, always lands up with a
'Text
to Speech'.
I use CW Skimmer to display the morse code ( I decode it visually) and have
successfully made many CW contacts. Now I want to try SSB.
Does anyone know of such software. SPEECH to Text - not text to speech!
Your help appreciated
John G4BOU
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|