Speech recognition technology has intrigued me for years but never seemed to live up to its promise. It always seemed to me to have more potential where a very limited grammar was involved, rather than attempting complex work like writing freeform prose. For example, the number of commands you can give to a television set is quite limited and much more likely to be successful.
Writing code in programming languages also seems to fit this quite well as it has a constrained grammar. I've often wondered what on earth I'd do if I couldn't type for some reason and figured that speech recognition might provide the answer (while hoping not to be in that situation in the first place).
I was impressed today to see a video from the folk at Renaissance in Israel (including fellow RD Jackie Goldstein) who have tried to tackle the use of speech recognition within Visual Studio. It's a work in progress but definitely worth a look:
http://www.renaissance.co.il/VSSpeech.aspx
Some customer support systems already use voice recognition, and I can never get them understand what I am saying. So all I can do is to say some expletive, clearly and slowly – that always works as expected, the system immediately says something like "Press zero to talk to an operator". Human operators always understand what I am saying…
Speech recognition can have undesirable unintended consequences. Around 1980, a vendor of computerized drafting workstations removed speech recognition from its product. One key customer had several drafting workstations in the same room. Many hours of tedious work was lost when one draftsperson said delete, which was heard at each drafting workstation. The customer was quite upset and the speech recognition code was commented out shortly thereafter.