[MV] Re: ViaVoice vs. iListen

T. Patrick Henebry tphenebry at comcast.net
Tue Jun 17 18:05:38 PDT 2003


>I guess the next logical question then would be,  is Windows voice
>recog software better than the Mac offerings in this point in time? 
>
>If yes, is it enough better to justify the purchase of a Windows
>machine?
>
>Let me be the first to type "that depends!" :-) because of course
>it's true that a persons individual situation is the key element
>of  the equation.
>
>Really I'm just hoping for a ballpark estimate such as,
>is Windows voice recog software 5% better, 25% better,
>50% better etc.  Or is it not actually better at all?

Speech recognition software for Windows has always been ahead of what 
is available on the Macintosh platform. In fact, the Macintosh 
platform has still not caught up to wear Windows was in terms of 
speech recognition back in 1999. I say this as someone who has been 
using speech recognition since the days of DragonDictate for DOS.

If you are someone who is using speech recognition due to a 
disability and needs a hands-free computing environment Windows, and 
NaturallySpeaking Professional are what you need. (ViaVoice Pro for 
Windows also allows total hands-free control of the computer, but it 
is a bit more cumbersome and does not to support a scripting 
language.)

Even if you were not using speech recognition due to a disability the 
Macintosh platform simply does not offer speech recognition programs 
with the combination of features available in Windows speech 
recognition software. For example, a physician or lawyer could find 
it very useful to have a speech recognition program which: supports 
digital recorders, allows you to save audio data so that you can go 
back and complete documents which cannot be finished in one dictation 
session, has tight integration with Microsoft Word, supports third 
party add-on vocabularies, and also allows editing and correction by 
voice anywhere text can be dictated. This combination of features can 
only be found on the Windows platform. What's more you have the 
choice of more than one application that offers this combination of 
features. Cost is another factor. Speech recognition is extremely 
demanding application in terms of both processors cycles and memory. 
If you need to buy a new or hardware keep in mind that you can 
purchase All of the necessary hardware and software to read 
Windows-based speech recognition (including a copy of Microsoft 
Office) for about the same, if not less, then the cost of a new G4.

Both NaturallySpeaking Preferred and NaturallySpeaking Professional 
offer the easiest correction and the best editing by voice of 
anything else on the market. So if you are someone who dictates 
several thousand words per day and also needs to perform extensive 
editing it could well be worth using a Windows machine. If you are 
going to be writing an extremely long document such as a book then 
NaturallySpeaking Professional is your best choice. It is the only 
program on the market that has the editing and correction 
capabilities of NaturallySpeaking coupled with the ability to save 
audio data. (ViaVoice Pro for Windows can save audio data too; but 
it's easier to correct and edit in NaturallySpeaking.)

If you use speech recognition we simply as a means to avoid typing 
(or are disabled but do not need a completely hands-free computing 
environment) then either ViaVoice or iListen will suit your needs. It 
then comes down to choosing and which one best suits the way you wish 
to work. Every speech recognition program, both Macintosh and 
Windows, has its own idiosyncrasies which must be dealt with.

I am not saying that you should not use Macintosh based speech 
recognition. But it simply does not measure op to what is available 
on the Windows platform. For some people it is worth the expense of 
purchasing a PC in order to run speech recognition, and for others it 
is not. You cannot really assign a simple percentage value.

Both iListen version 1.6 and ViaVoice for Mac OS X will allow way 
user with a well trained voice profile to dictate text at 
approximately 100 words per minute with a 98% accuracy rate. I say 
approximately 100 words per minute for two reasons: that sort of 
speed is only available with ViaVoice when using SpeakPad, and you 
must slow iListen down in some applications to avoid scrambling your 
text. ViaVoice does have that tendency to periodically go deaf, 
making it necessary to restart the program. If you first save your 
work, including the audio data (ViaVoice refers to this as saving the 
dictation session.) You will be able to resume right where you left 
off and still be able to correct or edit by voice. Assuming you were 
dictating into SpeakPad. Dictating into anything but SpeakPad when 
using ViaVoice his excruciatingly slow.  With ViaVoice you can only 
correct or edit by voice in SpeakPad. Only iListen supports digital 
recorders on the Macintosh. iListen will periodically crash on some 
people's systems when used with some applications. I am sure 
MacSpeech is working to resolve this. These days I find a myself 
using iListen unless I am working on a longer document, one in which 
have most frequently use either spelling the mode or text macros, or 
I must edit text that I have not to dictated but have instead 
imported from another source. (The presence of preexisting text can 
confuse iListen leading to scrambled text.)

ViaVoice has advantages over iListen in the following situations:

Documents which require the frequent use of text macros or text 
entered through the use of spelling mode. Currently iListen does not 
properly track text entered into a document by either of these means. 
This can lead to problems later when either editing or correcting 
recognition errors (basically the document text gets scrambled). 
There is a work around for this involving using the "Commit 
Corrections" command. However doing so the means you lose the ability 
to edit by voice any text dictated prior to the use of this command. 
It also means you must issue extra commands before resuming dictation 
since by default there will be no space between the previously 
dictated text and the new text; in addition to this the first word 
dictated will be capitalized. So for example, even with the new 
improvements, iListen would not be appropriate for the gentleman who 
wished to use speech recognition to dictate DNA sequences. And I 
myself do not use iListen if I am going to be dictating a document 
that frequently contains Unix commands since this requires either 
text macros or the use of the spelling mode.

Since iListen lacks the ability to save audio data it isn't 
appropriate for documents that you cannot complete in a single 
dictation session. Once you exit iListen (or issue the "Commit 
Corrections" command) all audio data prior to that point is lost. 
This means you could not go back at a later date and either correct 
recognition errors in the previously dictated text or edit by voice. 
Since I depend on speech recognition due to problems with my hands we 
ViaVoice is definitely a necessity in this situation.

A legal or medical professional can take advantage of specialized 
vocabularies available for ViaVoice. These vocabularies can both 
tremendously increase accuracy and shorten training time for those 
who use them. This definitely makes it worth considering the ViaVoice 
instead of iListen for such professionals. For information on these 
vocabularies see:

<http://www-3.ibm.com/software/speech/mac/osx/vocabs.html>

When it comes to command-and-control ViaVoice provides the user two 
options should they wish to create their own voice commands: 
keystroke macros (which do have some limitations) or AppleScript. By 
contrast iListen is entirely dependent upon AppleScript for creation 
of voice commands. This makes ViaVoice a better choice for those who 
want to create their own voice commands but do not know, or wish to 
learn, AppleScript.

Finally, a user may occasionally encounter program which simply does 
not work well with iListen. In which case ViaVoice is the only other 
option for speech recognition under OS X.
-- 
----------
T. Patrick Henebry



More information about the MacVoice mailing list