As I have mentioned, I have been writing a lot lately, and my wrists and forearms have been suffering as a result. I learned a long time ago that the best thing you can do for your RSI is minimize the repretion, so I keep a plethora of input options available, from my precious Apple ergonomic keyboard (useable via Griffin’s iMate adb-to-USB adaptor) to a couple of graphics tablets to differing standard-layout keyboards.

One input methodology has always been avoided on my part, however, because of my early, extended exposure to optical character regognition systems (I helped develop a specialized data input system for art-auction results in the very early 90’s, pre-Windows). The input stsyem is not OCR-based, but observing how OCR developers evaluate and make claims of “99%” accuracy made me highly skeptical of similar systems’ marketing claims.

In OCR, a factual claim of 99% accuracy can be made when a 99 of 100 characters have been accurately recognized and rendered. Of course, this sentence contains about fifty-odd characters. So 99% accuracy can mean one error in every sentence or two. Which makes it less useful than one might hope.

As voice-recognition systems entered the market, I watched as similar claims of accuracy were made, and concluded that voce-rec had to overcome the same hurdles that OCR had yet to master.

Finally, with the the release of IBM’s ViaVoice sometime in 2000, and after hearing my father describe the reasonably good results he has experienced with Dragon’s line of voice-recognition dictation software, I concluded it was time to keep an eye out for a good price on voice-rec sofware.

Boy, did I find a good price. But that’s a separate entry.

So far, I’ve found ViaVoice to be reasonably effective. I have yet to really put it to work, but will certainly followup on this.