Trying speech recognition, then giving up

Lee is a pretty typical new speech user:

In a fit of wanting to type faster, I looked into the latest version of Dragon NaturallySpeaking. After 30 mintues of googling, I've decided that speech recognition is still not ready for prime time. :-

He notes correctly that even a 2% error rate means something like 4 or more errors in a short blog entry. With some diligence, you can get the error rate down closer to zero, as people with RSI or other disabilities regularly prove. You'll need to invest in a high-quality microphone, spend significant time training the system and keeping it up-to-date, and getting familiar with the correction UI. We're working to make things better, but if you're serious about trying SR, please don't give up until you've really tried.