Re: What am I doing wrong?

• Messages sorted by: [ date ][ thread ][ subject ][ author ]
• Next message: D.N. Thomas: "CODECS"
• Previous message: Chuck J. Cronenwett: "CODECS"
• Maybe in reply to: Hasselberg, Stephen: "What am I doing wrong?"
• Next in thread: Hasselberg, Stephen: "Re: What am I doing wrong?"
Hey! Who's out there selling DNSD to the medical community? I recently sold
one dictation station. Next week will be installing 5 dictation system
stations, enrolling 7 physicians, and three NORCOMS. Who else is doing this
type of selling/training? Also,.....stay tuned for my release of a Topic
Builder for the medical community. The Blister.


----------
> From: Patricia Hawkins <phawkins@@connact.com>
> To: holland@@rattlers.net
> Cc: SHasselb@@cldx.com; voice-users@voicerecognition.com
> Subject: Re: What am I doing wrong?
> Date: Saturday, May 30, 1998 1:26 PM
>
> >>>>> "HSJ" == Holland St John <holland@@rattlers.net> writes:
>
> HSJ> Steve...
> HSJ> I have tried the 55k vocabulary, didn't change my accuracy. Under
many
> HSJ> conditions the 55k vocabulary my decrease accuracy, since the
program
> HSJ> has to make finer distictions between more words that sound similar.
>
> HSJ> My understanding is that the 55k option is primarily for
professionals
> HSJ> with very large vocabularies spanning many fields. I'm a doctor and
use
> HSJ> *lots* of words not listed in the usual 70,000 word printed
> HSJ> dictionaries. Still, the base vocabulary seems to accommodate my
usage.
>
> Well actually, now you mention it, I do get great recognition in my
> 35,000 words-model "technical writing" topic; and okay recognition in
> my 35,000 word-model "creative writing" topic; but I was having a
> whole lot of trouble with my "casual writing" (basically email) topic
> when it was 35,000 words; it's the only one I have actually rebuilt on
> the 55,000 word model, and now I think about it, I might try
> re-building the creative writing topic, or rather, building a second
> 55,000 word creative writing topic, and experimenting; but I think my
> technical writing vocabulary is narrow enough to use the 35,000 word
> model.
>
> I don't think it's actually the number of words, but which words
> Dragon chooses to sacrifice when it brings new words into the active
> vocabulary, since someone had had a good experience deleting all the
> proper names and abbreviations that they didn't want, in a 35,000 word
> model.
>
> I'm also making a practice of deleting a few unwanted words whenever I'm
> in topic editor, but I really don't want to take the time to delete a
> lot of them all at once.
>
> HSJ> Your comments about the custom words and phrases matched my
experience.
> HSJ> I could find no easy way to transfer the custom words and phrases to
the
> HSJ> new vocabulary. The single words were not really too much of a
problem,
> HSJ> they are picked up running the topic builder. The phases are more
> HSJ> difficult. I generated a txt file of phrases by reading the phrases
> HSJ> listed in the topic editor into a tape recorder, then listening to
the
> HSJ> tape and dictating the phrases into Nat's editor window (one phrase
per
> HSJ> line) and saving the file. Then I added the custom phrase file back
into
> HSJ> Nat using the advanced options of the topic builder. Works, but not
> HSJ> easy.
>
> HSJ> Holland
>
> HSJ> Hasselberg, Stephen wrote:
> >>
> >> Any comments on this suggestion? Will increasing my vocabulary size
improve
> >> my recognition accuracy? If so, how do I do it? I tried creating a
new
> >> topic based on "Base General English -- 55k". But when I did all of
my
> >> custom words were not carried over. How do I import them? Do I have
to
> >> re-train all of them again?
> >>
> >> Steve Hasselberg
> >> SHasselb@@cldx.com
> >>
> >> ----------
> >> From: Patricia Hawkins
> >> To: SHasselb
> >> Cc: voice-users
> >> Subject: Re: What am I doing wrong?
> >> Date: Thursday, May 28, 1998 6:54PM
> >>
> >> I found that my error rate dropped dramatically when I changed my
> >> topics over to using a 55,000 word model. I had been using the 33,000
> >> words model. I was getting errors on variants of common words --
> >> plurals, verb forms, etc.. Someone posted, pointing out that Dragon
> >> sometimes sacrifices common word variants when adding new words; his
> >> solution was to delete the many unused proper names and abbreviations
> >> in the vocabulary; mine was to rebuild my topics using a larger word
> >> model. I also find that I need to be careful to speak clearly.
> >>
> >> >>>>> "HS" == Hasselberg, Stephen <SHasselb@@cldx.com> writes:
> >>
> HS> Okay, what am I doing wrong? I have been working with
NaturallySpeaking
> >>
> HS> (Deluxe edition) for about four months now. My recognition does not
> >> seem
> >> to
> HS> be improving, and I certainly don't have the accuracy that many
people
> >> seem
> >>
> HS> to claim. I would welcome any advice that people have. Let me tell
you
> >>
> HS> about my system -- which the Audio Setup Wizard rates as "Average":
> >>
> HS> IBM 200 MHz Pentium with MMX
> HS> 64 MB RAM
> HS> 2 GB hard disk (partitioned as two 1 GB disks; drive C 500 MB free,
> >> Drive
> >> D
> >>
> HS> empty except for swap file)
> HS> Sound Blaster 64 AWE sound card (not Gold)
> HS> VXI microphone that came with NaturallySpeaking (I do no use the
"bias
> >> box"
> >>
> HS> that came with the microphone. That dramatically decreases
performance
> >> and
> >>
> HS> gives a much lower rating on the audio setup Wizard)
> HS> there is also an ethernet card on my system
> HS> Operating system is Windows 95 4.00.0970a
> HS> I have Natural Text and Natural Word
> HS> I have installed the patch to Microsoft Office
> >>
> HS> I almost always dictate into NaturalWord, rarely using the Dragon
text
> HS> editor. I don't believe performance was any better in Dragon than it
> >> was
> >> in
> HS> NaturalWord.
> >>
> HS> I've been through the basic training and the extended training
(multiple
> >>
> HS> times). I have read most of the training packages, including the
longer
> >>
> HS> ones like 3001, Dogbert and Dave Barry.
> >>
> HS> I have tried increasing the microphone volume, without much
improvement.
> >>
> HS> I have been building up a collection of files for the topic builder,
> >> which
> >> I
> HS> run every couple of weeks or so. I probably have about 200 KB of
files
> >> I've
> HS> dictated.
> >>
> HS> I have tried Rose's training program (reading and then correcting a
> >> magazine
> HS> article) and can't say I noticed the "big improvement" she says I
> >> should.
> >>
> HS> I think I am dictating correctly; speaking swiftly in long phrases
and
> HS> sentences, speaking clearly, but not over enunciating.
> >>
> HS> I have not tried the usual suggestions (rounded up the usual gang of
> HS> suspects) new microphone, new sound card, more memory. I was
hesitant
> >> to
> HS> shell out even more bucks. And since the Audio Setup Wizard rated my
> >> system
> HS> as "Average" (which is really quite good) I didn't think it would be
the
> >>
> HS> microphone or the sound card.
> >>
> HS> In dictating this note I have accounted 27 errors (make that 28). Is
> >> this
> HS> typical? To my expecting to much? (Make that 29). While I'm still
> >> quite
> HS> impressed with the technology, it will take a big improvement for me
to
> HS> agree with Dragon's statement that it's actually faster than typing
> >> since
> >> I
> >>
> HS> have to do all of this correcting. With my RSI as bad as it is I'm
> >> trying
> HS> not to touch the keyboard at all, and this is not helping.
> >>
> HS> Well, Dragon just gave me an "unexpected exception" and terminated on
> >> me.
> HS> (I wonder what exceptions are expected?) I am finishing this by
hand.
> HS> This doesn't happen often to me, that Dragon crashes in the middle of
> HS> something, but it OFTEN refuses to exit. The disk thrashes for 15
> >> minutes
> HS> or more after exiting the program and I finally have to
ctrl-alt-delete
> >> and
> >>
> HS> End Task. Sometimes that doesn't even work and I have to reboot.
> >>
> HS> Please help. I'm trying to make this work...
> >>
> HS> Steve Hasselberg
> HS> SHasselb@@cldx.com
> >>
> >> ------ Message Header Follows ------
> >> Received: from smtp.cnp.cldx.com by msmail.cnp.cldx.com
> >> (PostalUnion/SMTP(tm) v2.1.8d for Windows NT(tm))
> >> id AA-1998May28.190019.1268.737697; Thu, 28 May 1998 19:00:19 -0400
> >> Received: from fw.cldx.com (fw [151.41.2.4]) by smtp.cnp.cldx.com
> >> (8.6.12/8.6.10) with ESMTP id SAA18400 for <SHasselb@@cldx.com>; Thu,
28 May
> >> 1998 18:49:06 -0400
> >> Posted-Date: Thu, 28 May 1998 18:49:06 -0400
> >> Received: from localhost by fw.cldx.com id SAA28177; Thu, 28 May 1998
> >> 18:38:55
> >> -0400 sender phawkins@@connact.com for <SHasselb@cldx.com>
> >> Received: from connactivity.connactivity.com(206.34.200.2) by
fw.cldx.com
> >> via
> >> smap (g3.0.1)
> >> id sma028175; Thu, 28 May 98 18:38:47 -0400
> >> Received: from WATERCAT.watercat (ppp-215-21.ConnActivity.com
> >> [206.34.215.21])
> >> by ConnActivity.ConnActivity.com (8.8.5/8.7.3) with SMTP id SAA26350;
Thu,
> >> 28
> >> May 1998 18:43:22 -0400 (EDT)
> >> Date: Tue, 26 May 1998 16:52:04 -0400
> >> Message-ID: <6219-Tue26May1998165204-0400-phawkins@@connact.com>
> >> X-Mailer: emacs 19.34.6 (via feedmail 7 Q)
> >> From: phawkins@@connact.com (Patricia Hawkins)
> >> To: SHasselb@@cldx.com
> >> CC: voice-users@@voicerecognition.com
> >> In-reply-to: "SHasselb@@cldx.com"'s message of Tue, 19 May 1998
15:51:47
> >> -0400
> >> Subject: Re: What am I doing wrong?
>


• Next message: D.N. Thomas: "CODECS"
• Previous message: Chuck J. Cronenwett: "CODECS"
• Maybe in reply to: Hasselberg, Stephen: "What am I doing wrong?"
• Next in thread: Hasselberg, Stephen: "Re: What am I doing wrong?"