zooming, predictive text entry!

"Dasher: Attack of the killer alphabets"

"Dasher is a zooming interface. You point where you want to go, and the display zooms in wherever you point. The world into which you are zooming is painted with letters, so that any point you zoom in on corresponds to a piece of text. The more you zoom in, the longer the piece of text you have written. You choose what you write by choosing where to zoom.

To make the interface efficient, we use the predictions of a language model to determine how much of the world is devoted to each piece of text. Probable pieces of text are given more real-estate, so they are quick and easy to select. Improbable pieces of text (for example, text with spelling mistakes) are given less space, so they are harder to write. The language model learns all the time: if you use a novel word once, it is easier to write next time."

Tags: , ,

8 Responses:

  1. brad says:

    Wow! It's so incredibly easy to use! These people should win a bunch of awards for accessibility/usability.

  2. flipzagging says:

    The predictive nature makes it a cool dada generator too. And it's twitch-interactive so who knows what subconscious things could be happening as you choose letters?

    take it was the universelfdestrong ust all thing blood the bit chured by dally needed two twiga head held the convership you could force thus isinxcuse ges of tengealer eive in cone ofqually mother in its with a betwere z and as selectrong about on to othere is a framnd lostphydrogen men x sort from being such brought nd wenty conclusion he lazilyrics affortly that is to the depoison of out of zero use

    This is my new personal oracle!

  3. waider says:

    Whoopee. One more goddamn widget library to download before I can build the damn thing. Still, at least they've got the source out where we can all poke at it, which they didn't have when it showed up in the usual meme pools last year.

  4. jerronimo says:

    I remember trying this out about a year ago, on a windoze application. it was pretty damned neat. Even for words that it didn't know, it was very easy to use.

    I could see it being MUCH faster than handwritten entry with practice.

    very neat tech.

  5. hfx_ben says:

    An old VRML PoV stunt was to fly through JPGs ... quite shocking to pop through a mountain and be confronted by _____________.

    I just finished plowing through a 650KB bookmark file (some dating back to 3Gold) ... knowledge space of another kind!*

    It's been a while since I've read Heidegger's Question Concerning Technology (techne is such a far out concept) ... I'm not sure we've decided if we discover knowledge or create it. (Does anyone talk about "noosphere" beside de Chardin and Raymond?)

    * Moz 1.3a choked twice, but only non-destructively. It /really/ hates dupes.

  6. evan says:

    I think this was announced on <lj user="sweetcode"> a while ago, which I heartily endorse (though lately it hasn't been so good).

  7. cryllius says:

    I just spent about 20 minutes in that, it's just about the coolest thing ever.

    You REALLY need to do another installment of free association using this!