The problem with voice recognition is that we tend not to speak using standard, perfect language.

Depending on where you’re from in the U.S., the brown, fizzy, caffeinated beverage you drink from a can might be called soda, soda pop, pop, or “Coke,” even if what you’re drinking is a Dr. Pepper. And if a common word is used, it might not sound the same. “Caramel” is pronounced “car-ml” in Colorado and “carra-mel” in Maine.

Historically, voice recognition systems haven’t done well with accents or dialects, but that’s finally starting to change.

Tmall Genie, Alibaba’s digital assistant, is now being trained to recognize some of China’s many regional dialects. Amazon’s Alexa is similarly being trained to recognize accents and dialects as well as ethnic origin.