DARPA Project Babylon: Universal Translator

silance writes "Take a look at this project from DARPA (Defense Advanced Research Projects Agency)! This time the boys are trying to hammer out a portable, two-way, real-time, multi-lingual audible speech translator proposed to be run on everything from PDA's to wearable military hardware to workstations (to replace their PRE-EXISTING ONE-WAY real-time hand-held audible translators, of course!). The site contains descriptions of technical approaches, a technical milestones timeline, and a nifty Power Point presentation for the executive-types ;) They should give William Shatner a beta model out of pure respect... Here's a link to Google's cached HTML version of the Power Point presentation just in case. (P.S. - get a load of that logo at the bottom of the page!)"

Comments

Re:Give it to Hoshi

[uebernewby]

True. Hoshi's fiddling with the universal translator really made me think about that piece of equipment we've been taking for granted in previous Star Treks.

Seems my university syntax and phonology courses weren't *that* useless after all...

The way I see it: suppose Chomsky's Universal Syntax turns out to be not innate to human brain structure, but to the very essence of communication. Meaning: if you're going to communicate something, all the forms you're going to be able to do it in will conform to a fairly basic set of ground rules and all the intricacies of natural languages are simply icing on the cake, as it were. If you figure out what that Universal Syntax is (sorry, I forgot the exact term he used - it's been a while, and my university education was in Dutch), you can feed that into a computer and teach it to reduce all phonemes from a given language to it. Then you can have the computer expand the basic message back into coherent communication in another language using the same basic rules.

It's late. And when it's late, this is the kind of stupid stuff I think about.

Oh, and I don't think Hoshi's *that* cute.

N-Gram

[rufusdufus]

For those who don't know, an N-gram is data structure which encodes the statistics of word order in a language. These are used to greatly improve the accuracy of language pattern matchers such as speech recognition.
A typical speech recognizer might use a 3 word N-gram (tri-gram), which keeps track of all probable words which follow and thier likelyhoods. The probabilities are calculated by running terabytes of english text (books, magazines, internet chat boards) through a word counting program.
Thus, "green eggs and" will get a very high probability for "ham", but low for "jam", so it can bias a sound that seems to match "mam" acoustically to the more likely linquistic match "ham".