[Theo]

There is voice recognition software that'll take into account different accents. Nuance, for example, has one.

I'd blame this problem on the developers, who should have known that the "users" would have accents....

Ted, a Voice User Interface developer....

[Myrddin]

There is voice recognition software that'll take into account different accents. Nuance, for example, has one.

Nuance is generally good. I've worked with Conversay and the new reco engine from Vox Industries as well. The midwest speakers have 100% recognition. I've experienced major problems with regional accents. A customer from Boston says "Fah wood" in place of "forward". Her Swedish collegue says, "farrr vard" where the "rrr" indicates a rolled "r". IBM claims to have their "ViaVoice" adjusted to handle all U.S. regional accents. The toughest task is reco on a Japanese or Korean native speaker with a heavy accent. It's almost hopeless.

The text to speech engines have some issues too. The first pass speaking Pocatello was rendered "poh kay' tel oh". Wapakoneta was rendered "wah pay kon net uh". Many of those gaffes can be fixed with phoneme/diphone/allophone type symbol sets to construct a proper rendition.

[Blood of Tyrants]

The problem with that is developers generally use their own voices (and the few around them who tend to talk like them) to test their software.