Hello EASI colleagues,

I would like to share some of my impressions after having upgraded from
version 4.00 to 5.00 of NaturallySpeaking. I conducted my tests on a Dell PIII
550 desktop with 256 MB RAM, and a Dell PIII 800 laptop with 192 MB RAM.

Version 5.00 is the first NatSpeak upgrade since L+H purchased Dragon Systems
last year. Version 5.00 is a significant upgrade, and apparently represents an
attempt to merge the best of VoiceXpress with  the best of NaturallySpeaking.
NaturallySpeaking was, in my opinion, a much better product. VoiceXpress, for
example, has few editing commands, whereas NatSpeak has many ingenious
commands for navigating around and revising documents. Although Version 5.00
has much in common with Versions 4.00, 3.52, and 3.00, it also seems to have
inherited some of the design philosophies that made VoiceXpress problematic.
So while there are notable enhancements, there are also features and operating
characteristics that may render the program less usable, both to people with
and without disabilities.

My overall assessment: Dragon Systems NaturallySpeaking version 5.00 is a
mixed bag. Consider carefully before you upgrade, and decide whether the
enhanced functionality is worth the functionality you will sacrifice.

Accuracy: I detect no significant improvement in accuracy between the two
versions. If you have achieved good accuracy using version 4.00, you may well
be satisfied by what you already have. A warning: When upgrading to version
5.00, I recommend that you do not bother allowing the program to modify your
old voice files for use in the new version. When I allowed NatSpeak to modify
my existing voice files, accuracy was much worse than when I re-created voice
files from scratch. (There are two utilities, "GetWords" and "PutWords," on
Joel Gould's unofficial NaturallySpeaking web site that make it easy to
transfer vocabulary and pronunciation from one voice file to another.)

Correction Dialogue: I am extremely disappointed by changes in the Correction
window. In version 5.00, you can choose between two different Correction
windows. (I have not yet worked with the new "Quick Correct" technique because
I initially found it too restrictive.) Unfortunately, the fonts in both types
of correction windows are hard-coded in a tiny, spidery, 8-point typeface, and
the information in these windows is difficult to see. In version 4.00, it was
possible to adjust the Display Properties Applet in the Windows Control Panel
to increase the fonts size in the correction window. (I believe that the
"Message Box" item affected the appearance of the NaturallySpeaking correction
dialogue.) This is a major usability/accessibility blooper. Despite the fact
that I have "perfect" corrected vision, I am forced to crane my head forwards
to read the correction dialogue, even with a 17-inch monitor. On my laptop,
the correction window is even less legible. Furthermore, "Automatic Playback
on Correction" does not work as well as in version 4.00. More often than not,
playback reads an extra word at the end of the phrase or word you are
correcting. Thus, visual and audible feedback do not correspond, which is
confusing and can lead the user to make mistakes during correction. In the new
version, the user must wait until automatic playback is complete before making
corrections by voice; in version 4.00, the user did not have to wait -- you
could issue commands at any time. Consequently, correcting errors with
automatic playback switched on can be a source of error, and I have had to
disable the automatic playback feature. In addition, after making a choice
from the Correction window, the new version is more likely to ignore the rules
of capitalization than the earlier version.

Mouse-free operation: There are a number of new usability problems related to
the keyboard-only interface. Unlike Version 4.00, a mouse is required to
perform several steps during enrollment. In fact, the entire program is
somewhat more mouse intensive than in the past, and the keyboard interface is
less intuitive and awkward to use than in version 4.00. The awkwardness of the
keyboard-only interface can translate into ungainly voice commands. For
example, in certain situations you must say something like "Press Keypad Star"
to put focus to the "DragonBar," which is a floating toolbar that contains two
NatSpeak menus and several (unlabelled!!!) toolbar buttons.

Status line information: In version 4.00, the status line in NatSpeak's
proprietary editor provided useful information about recent utterances. Some
of this information is no longer available. The only way to get any of it is
to display the Results Box at all times. However, the Results Box is visually
distracting, especially for users with certain learning disabilities. I have
had to turn it off for many of the people who I have trained; personally I
have never seen a need for it. In version 5.00, hiding the Results Box means
that users cannot obtain as much information about the state of the program.
Even with Results Box showing, the program no longer reports events such as
when a "Select" command fails to find text.

Compatibility: This is the good news. Version 5.00 works more seamlessly in
more applications than version 4.00. I have found that I can reliably dictate
into Microsoft Word and Outlook without significant loss of accuracy or speed
or features. Note, however, that Version 5.00 appears to be at least as crash
prone as its predecessor when dictating into applications other than the
Dragon Systems proprietary editor.

Activating menus, and dialogue/toolbar buttons: More good news. It is no
longer necessary to say "Click" to activate menus and dialogue buttons. There
are situations, however, when not uttering "Click" can give unexpected
results. For example, while inserting the word "Insert" into a document, the
program will pull down the "Insert" menu! But this is a minor inconvenience,
and one of the annoyances that one must accept when operating a PC by voice.

New features: There are noteworthy new features: It is now possible to create
multi-line voice macros in the Preferred Edition. (In the past, you were
limited to one line voice macros.) Canadian and British users will appreciate
the ability to insert properly formatted postal codes directly by speaking.

Overall assessment: If you are using version 4.00 and are happy with it,
consider carefully before upgrading. For my part, I will continue
experimenting with the new version, but I don't think that I will give up on
version 4.00 yet, especially on my laptop. The best thing about version 5.00
is that it is compatible with more programs than earlier versions. The worst
thing about version 5.00 is that correcting misrecognitions is horrendously
more difficult than in the past. I consider the ease with which one can fix
misrecognitions as a rough measure of the overall usability of a speech input
product. The easier it is to correct misrecognized utterances, the more likely
it is that a user will be able to tune the voice files for accurate and speedy
dictation.

I will be presenting a paper on speech recognition as educational and
workplace accommodations at the CSUN conference in Los Angeles in March. For
more information, see:

www.interlog.com/~acantor/upcoming.htm

Alan

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