Several months ago, I upgraded my voice recognition software to Dragon NaturallySpeaking version 8. I have been using voice recognition software for over five years now, and have been very satisfied with it, although older versions could be maddening at times–and hilarious at others. This version is amazingly accurate–my only complaint since upgrading is that I am not having nearly as much fun proofreading my notes as I have the past.

Voice recognition software is rather remarkable technology, and to watch the medical version crank out complex terms with aplomb is almost magical. The software does not make spelling errors, but contextual errors are relatively common, and need to be carefully screened before sending your documents out the door.

A few of the contextual swaps I have encountered are as follows:

What I said: The patient said Viagra wasn’t working, and wanted to try Cialis.

What it typed: The patient said Viagra wasn’t working, and wanted to try and see Alice.

What I said: The patient was thoroughly reevaluated by her previous physician.

What it typed: The patient was thoroughly violated by her previous physician.

What I said: The patient will be started on depot testosterone.

What it typed: The patient will be started on devil testosterone.

What I said: The patient has prepubertal pubic hair changes.

What it typed: The patient had puréed pubic hair oranges.

The accuracy of this latest version has made such–ahem!–interesting errors quite a bit less common — to the point where I was becoming a bit complacent about careful proofreading. That is, until today:

What I said: The patient was seen several years ago for problems of urge incontinence, and has been on Detrol with good symptom relief.

What it typed: The patient was seen several years ago for problems of urge incontinence, and has been on a troll with good symptom relief.

And you thought they spent their entire lives living under bridges…