Open text transcription over the phone is still a pretty dodgy affair. Take Google Voice: the technology behind it has been applied, by Google, to many of its projects including its map product for mobile devices. And sometimes the results are spectacular; I can pick up my phone, fire up Google Maps, press a button, say the location I want found, and Google will, under ideal circumstances, find the restaurant/bar/retailer/point-of-interest in a few seconds and then send me directions. But just as often, it fails miserably by sending me directions to a site in Connecticut when I’m really just looking for a restaurant 3 blocks where I may have been standing in downtown Boston.
But the one space where this kind of near-but-not-quite real time open transcription is with text messaging. Not that I’d encourage anyone to pick up their phone while driving, but if you’re going to pick up your phone, you really shouldn’t be sending text messages. And, fortunately, many states happen to agree with that sentiment, having banned text messaging while driving. But what if you could send a text messaging by speaking the message that you want sent? Would you be willing to put up with 80% accuracy if you had the ability to send a text message while driving without a) taking your eyes off the road or b) getting fined in those municipalities where texting-while-driving is illegal?
We tried building an IVR application that does just this recently. The app answers the phone, asks you for the destination for your text message, then it asks you to record your message. Two to three minutes later, after it’s been sent via SOAP to be transcribed by an automated transcription service, your intended recipient gets a text message. This is all fine and well…but we still have a few things that would need to be solved:
1. Since the transcription accuracy is around 80%, it’d be nice to actually see or hear what actually got sent. Unfortunately the transcription is still near real-time. So unless you want to linger on the phone for several minutes while the gears turned at some distant outsourced automated transcription service, there’s no way to do this short of having the IVR call you back or, worse yet, sending you a text message which you really really should not be reading while driving.
2. Speaking of receiving a text message, usually a text is part of an ongoing conversation. So how would you get the reply message? By text? That would defeat the purpose of this application. By phone? That’s better, but there, again, is a limitation because…
3. …our SMS provider is outbound only. So your recipient can’t get a message back to you anyway, because if they reply, they’ll end up sending a message back to a shared short code rather than to you.
We might be able to solve item 3 soon. And, if so, we’ll back to report on progress.
