EASI Archives

Equal Access to Software & Information: (distribution list)

EASI@LISTSERV.ICORS.ORG

Options: Use Forum View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Alan Cantor <[log in to unmask]>
Reply To:
* EASI: Equal Access to Software & Information
Date:
Wed, 16 Apr 2003 13:55:58 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (123 lines)
Hi Bill,

Your voice files may be corrupted. Solution: Create a new user.

The job of creating a new user is not as daunting as many people imagine. The
"hard" work is the preparation. If you take time to prepare, you will be up
and running in short order. You may be surprised how quickly you achieve the
accuracy you are accustomed to. I'm talking hours, not weeks.

Here are three techniques to speed up the process of creating a new user:

Technique 1: This is optional. Use the GETWORDS utility to preserve your
unique words and expression, and the PUTWORDS utility to transfer the data to
your new voice file. The utilities are found on Joel Gould's unofficial web
site:

http://www.synapseadaptive.com/joel/getwordsautilityaprogram.htm

Getwords extracts a list of words and their pronunciations from your current
voice files and saves them to a text file. Putwords inserts the list in a
different voice file. (The utilities are built into NaturallySpeaking 6 and 7;
but version 6 has a lot of problems; and if you switch to version 7, you may
be disappointed in how slowly it responds compared to version 5, with little
or no improvement in accuracy. Be skeptical of Scansoft's marketing hype that
version 7.0 is 25% more accurate than version 5. If Version 5 is working for
you, there is little incentive to upgrade.)

You do not have to use these utilities. I do not use them anymore. In version
5, there are two other techniques that work about as well -- and maybe
better -- to jump-start the creation of a new user:

Technique 2. Create a list of your unique words and expressions. Then use the
NatSpeak's 5.0 Vocabulary Builder to add the list to the new user. (This is
Step 1 in Version 5's Vocabulary Builder.)

When creating the list, put each word and expression on a separate line, and
save the list as a plain text file. The file might look like this:

EASI
Project EASI
Bill McQueen
William McQueen
colour
colours
labour
labours
laboured

etc.

It sounds like a lot of work, but in fact, creating such a list is trivially
easy. The custom dictionary from your word processor already contains dozens
or hundreds of these words. Copy these words into a new file, and add a
handful of expressions and names. I did this at one point, and I think it took
me all of 10 minutes.

Technique 3: Gather together several word processing or text files that
contain samples of your writing, and use the Vocabulary Builder to analyze the
documents. (This is step 2 of the Vocabulary Builder.)

Hint: Run a spell checker on these documents before handing them to the
Vocabulary Builder for processing.

Here is my "recipe" for creating a new user:

1. Optional: If voice files already exist, use getwords to extract unique word
and pronunciation information.

2. Create a list of your unique words and expressions, as described above.

3. Gather together a bunch of documents that contain samples of your writing,
as described above. (Spell check them first.)

4. Create a new user. Run through the Audio Setup Wizard and do the initial
reading/training (about 10 minutes).

5. Do Step 1 of the Vocabulary Builder to add your list of words and
expressions, and Step 2 to analyze your documents (about 10 minutes).

6. Optional. If you used getwords to extract word/pronunciation information,
use putwords. (5 minutes).

7. Do another training session. (NaturallySpeaking | Advanced | Train User...)
(about 25 minutes)

8. Dictate for a while, and be assiduous about correcting misrecognitions. (1
or 2 hours).

At this point, you should find that accuracy is about as good as before. In
fact, it may be better. Beyond a certain point, voice files do NOT improve.
After a few hours or days of correcting misrecognitions, gains in accuracy are
due more to adding new words and expressions. But if you use the techniques
outlined here, you will have taken care of much of this work.

Alan

Alan Cantor
Cantor + Associates Inc.
Workplace Accommodation Consultants
[log in to unmask]
www.interlog.com/~acantor


> I have been using Dragon NaturallySpeaking Preferred version 5 in
> WIN95B since I bought it in 2000. (A separate version which was
> not included with WP9) .... The programme has worked well,
> flawlessly, I should say, UNTIL last week. (I know the box doesn't
> include WIN95B, but several other experts correctly said yes it
> could support v5.)
>
> I have twice tried to "repair" it, and lastly, removed it (saving my
> speech files) and reinstalled it with the older speech files.
>
> It isn't very long before the programme undertakes an abnormal
> termination, and the following error messages have been received.
>
> SDAPI Error 4: Error: [voc\wordlist 4] SDWord_GetName: 55878
>
> Errors in Microsoft C++ Library
>
> NATSPEAK caused an invalid page fault in module DMFC6.DLL at
> 017f:5f442dbc.

ATOM RSS1 RSS2