Speech-to-text: you can't handle the truth!

"Dude be ready" the text message said. Probably. I didn't send it or receive it, I was just sitting next to my wife as she spoke it.

She was using her Droid's speech-to-text capabilities to avoid having to type while she drove to pick someone up. Now the Droid is awesome, but I'm pretty sure the "OR ELSE" I heard in her tone of voice didn't make it into that text message.

Sigh. Yet another way to strip nonverbal cues from our communication.

Every time I see another example, I'm still amazed that the march of technological progress doesn't carry forward all the information content of the old technologies it supersedes. A brief list:
  • Music with lossy compression (flogged here)
  • Speaking face-to-face -> telephone -> text -> speech-to-text
  • Cameras with fixed focus and tiny lenses
Each new technology is more convenient than the last, or makes inroads into new applications. But if computers were like this, your new Windows 7 box wouldn't recognize thumb drives because, hey, you've got online storage now.

Maybe we don't want the details: the closed posture, the reddening complexion, the tightening voice. The kick drum behind the vocalist. The beginnings of a dry brown edge on a flower petal. Maybe we'd say if it doesn't fit on one side of a sheet of paper in doublespaced Courier 12, it hasn't been edited enough yet. Maybe we want plausible deniability. The opportunity to be inhumane through ignorance.

You can't handle the truth! A fleck of spittle lands on Tom Cruise's cheek. Jack Nicholson hasn't had a cigar yet today.


  1. Well, you know how I feel about some of this. A low bandwidth channel can force people to be more clear and explicit instead of letting the umm-ing and pausing and vocal tone do the communicating. I have no more to say than what I just did, and it required from all of us the minimum of effort.

  2. Well, yes, force people who are in the habit of thinking about how they're heard. You just finished writing instructions on how to build an amplifier; imagining all the possible incorrect ways your words could be read was absolutely vital. Many people don't have that habit - witness the example in the first paragraph. For them, a low-bandwidth channel causes them to imagine they're communicating something quite rich when often they're not.

  3. You know, I'm halfway behind this idea. I like the idea of being able to dictate text for cell TXT messages and email but I hate having text read to me by a robot voice. I like reading, but on many devices typing is very inconvenient. There are services that will read your email to you over the phone. I can't imagine why anyone with working vision would want to turn a random access medium like text into a linear medium like speaking. But the other way around, OK.