My friend Steve and I were talking today about accent reduction software. Now, when you Google for that, you see a bunch of boxed products, like that only run on Windows. What Steve and I mean when we talk about it is more of an open library of software, on which enterprising people can build for their own uses.
As we settled on a general idea, he asked “Who’s doing this?” I thought of Google, who is clearly doing some good work right now with both speech synthesis and speech recognition for their Google Now service, but as far as I can tell there is no Google Accent Reduction. Same with Apple: can you tell Siri to point out when you say things with a Chinese accent?
What Google does have, I’m sure, is a huge sample of voices saying the same words in different contexts. If they wanted, they could be the authority on how to pronounce any word in any city, according to how the average resident of that city pronounces it. (I expect that feeds into their speech synthesis algorithms too.) For whatever reason, though, it doesn’t seem that they want me to have this data.
Steve wondered out loud, “Who could I get to start this…?” It hit me almost immediately: “Randall Munroe! The guy who does xkcd!”
Randall had already done this for his colors project, which is really similar from the perspective of data gathering, except that now he could also combine his other pet project, Basic English. Imagine the data he could gather if he asked all his readers to pronounce a selection of the “ten hundred” most commonly used English words!
He has the reader base, the technical acumen, the experience, and the drive to do something like this, and he would open the data as a SQLite zip file or whatever at the end of it all. Okay, maybe he’d have to brush up on his HTML 5 audio capture skills, but doesn’t he do this kind of on-the-fly learning anyway whenever he wants to do something especially interactive on an xkcd strip?
Steve’s final comment was to the effect of “Whoever [solves the problem of helping billions of people to speak more effectively to other people throughout the world] should win a Nobel Prize.” Hell, I think Randall should win something for that colors thing, but yeah, this might benefit more people.