By: Gary S. Kearney

Efficiency is getting the job done //right//. Effectiveness is getting the //right job done//.

VR (Voice Recognition) software claims to be a fast, easy and accurate way to turn speech into text. Users can dictate into virtually any application with benefits for:

* Professionals who create documents.

* Disabled PC users looking to optimize productivity with their PCs

* Home users who want to have fun with their computers. (See the goofs at the end of this article.)

* Speeding up routine tasks on the PC, such as creating documents, entering data, launching applications, sending e-mail, completing forms, and browsing the Web. You may complete all of these tasks at up to three times faster than typing - by speaking at up to 160 words-per-minute! (Note: Some of these options are only included in more expensive versions of the software.)

* Increasing productivity when away from the office by transcribing recorded dictation from mobile devices, including Microsoft Pocket PC and commercial digital recorders. (This is only available in higher priced versions.)

* Helps to protect from Repetitive Strain Injuries (RSI).

These claims overstate the situation a bit. The software is fairly accurate, but the type of errors produced are difficult to detect because the software produces text with correct spelling and grammar, which means that your spell checker and grammar checker will NOT find the problems for you. See the jokes at the end of this article for some examples. Some programs will read the text back to you and this is probably the best way to check it yourself, as the eye tends to see what it expects to see while reading. Any article intended for publication should be checked by an independent reader before it is submitted. Most VR software requires some training to work well, and will improve in accuracy as you use it. Using a special noise canceling microphone and taking pains to reduce the extraneous noise while you are recording will also pay off with increased accuracy.

The price and system requirements will change depending upon the features that you require. The least expensive options provide only for the recognition software itself, and the most expensive include: a noise canceling microphone, ability to transcribe text from digital voice recorders and other files, along with other special services such as using your voice to control your computer and replace typing as input to other programs.

IBM Via Voice: http://www-3.ibm.com/software/speech/

IBM has been involved in speech recognition for more than 40 years. Their Via Voice VR software has always been highly regarded. They provide a number of different options ranging in price from $60 to $200 for the home user. The more expensive versions include a microphone that would cost $40 if purchased separately.

Dragon Naturally Speaking: http://www.scansoft.com/naturallyspeaking/

Dragon Naturally Speaking software competes with IBM. The rights to the software have been recently purchased by Scansoft, and they are the current publisher. The features, prices and options are quite similar to those offered by IBM. The Sony ICD-BP150 digital voice recorder is also available as a bundle with Dragon Naturally Speaking voice recognition software for $250.

Computing Out Loud: http://www.out-loud.com

Susan Fulton's independent web site is devoted to speech recognition. She has tried to make her site as useful as possible for people who are investigating speech recognition for whatever reason, including repetitive strain injury. She is not promoting any single product, and is not a vendor. Her chief goal is to be a jumping-off point to other sites and to provide practical advice. Check out the goofs to grin at section where I found most of the examples shown below

Speech Recognition FAQ: http://www.tifaq.com/speech.html

This FAQ covers speech recognition systems and related topics including manufacturer descriptions of their products, where they can be purchased, and comments from users of these systems. Please visit manufacturer and reseller links for current prices and system requirements.

Since this software is quite expensive it's worthwhile to check out other sources such as used computer stores, garage sales, online auctions, or even the local thrift store or Salvation Army. One thing to be aware of is that some older versions of VR software use what is called "discrete speech" as opposed to "continuous speech" accepted by the newer versions. You are required to learn to pause slightly between words to allow the program to find the ends of words. This will slow down dictation slightly but is not really difficult once you learn it. The accuracy is more important, since greater accuracy will reduce the amount of time needed for corrections. There have been and are a number of other vendors for this type of software, but many of them have gone out of business and others are not well known. I would advise sticking to the two major vendors unless you get a very good deal.

If you wish to use voice recognitions software to transcribe notes from your recorder, you will need to be sure that the program you purchase will import the output files from your recorder. It's probably a good idea to check with the maker of the software to see if they recommend any particular hardware for use with their software; or conversely, to check with the software maker to see if they have any recommendations for a digital recorder.

Other Operating Systems:

IBM Via Voice is available for both the Mac and Linux platforms as well as the IPAQ; however, the latest version of Mac OS X includes all the software needed for VR and text to speech.

Check the specifications online or on the back of boxes in your local store to decide which features are most important to you. If you can live without certain features, you may be able to save money by getting a less expensive version. If at all possible, find someone who owns the software so that you can see how it works before you lay out all that money. Check with a local computer club if you can find one. If you have a student in the family it is also worthwhile to check out educational discounts.

Speech recognition software generally requires a rather powerful system, so pay careful attention to those requirements. If your system is close to the low end of the ones supported, the program is likely to run slowly. It can be annoying to have to wait on the machine, but most newer systems should work fine.

You may be able to return software if you find that it does not fit your needs, but don't depend on this. Such guarantees are subject to change and revocations. You may also be responsible for restocking or other charges. It doesn't hurt to ask though.

I have used older versions of both Dragon Naturally Speaking and IBM Via Voice software. I found that both had good and bad points, but I didn't use either one enough to make them viable. I found that the time I had to spend correcting errors pretty much offset the time I gained by faster text entry. I expect that the newer versions would do much better. If you are a good typist, there isn't too much benefit; but if you have problems VR could be the answer.

If I were to decide to purchase a digital voice recorder, I would definitely get software to transcribe my voice notes as that would greatly increase its usefulness: however, I would not usually expect to use the software to transcribe recordings of other people.

A good noise canceling microphone is probably worth the money, and you really do need to turn off the radio while recording. Don't expect the software to do much with speech recorded in a crowded airport terminal.

Many times the mistakes made by VR software are rather amusing. I publish a family newsletter where my software changed my mother's "standing ovation" to a "standing violation." Believe me I had a lot of feedback on that mistake.

Here are a couple of goofs from Larry Rosenthal along with a couple of things he's done to ameliorate the problem:

To a restaurant critic: "The food was delicious, the service superb, but I did not care for the ambulance."

To the credit card company: "I am writing again concerning the fourth item on your July statement ($345, Beale Jewelers). I assure you I am certain they never sold us a pair of herrings. My wife does not even wear herrings."

Larry reports that he programmed his VR software with a voice macro named "you %*\#@^&!." Now, whenever he notices that the screen contains a word of dubious origin, he roars, "you %*\#@^&!," and the software responds, "I'm sorry I was such a jerk. I swear it'll never happen again. I'm feeling much better now."

He now appends the following to all documents created with his voice recognition software:

James Richards gives this goof:

Podiatrist dictates: "patient presents with an accute tedis pedis..."

Program types: "patient presents with a cute tiny penis..."

The Podiatrist reported that he was glad he was proofreading that day...!

Margaret Plenert gives us this:

The intended text: "We'll need at least one "extrovert per group."

The program came out with, "We'll need at least one extra virgin per group."

Allan Stokes gives this one:

"We just finished a large turkey dinner and I'm expecting to see you shortly."

And what the software inferred: "... I'm expecting obesity shortly"

Personally, I think his personal computer was getting a little too personal on that one.