The military is looking for a better way to cut through the challenges of listening to radio chatter in several dialects when every word and second counts.
Currently, the military passes around the radio to interpreters who piece it together but lose a lot in the translation.
Darpa’s is experimenting with a system dubbed Robust Automatic Translation of Speech, or RATS.
Darpa wants RATS to pull speech out of “noisy or degraded signals,” with 99 percent accuracy at distinguishing spoken words from background noise. The system should also be able to identify, with 98 percent accuracy, the language spoken — with special emphasis on Arabic, Farsi, Pashto, Dari and Urdu.
The RATS software will be programmed with voice-recognition technology, to identify people on a military most-wanted list. It’ll also be able to automatically detect specific, preselected “key words or phases.”
They want the program to translate 15 languages among 1,000 different speakers with the ability to identify 100 words in three Middle Eastern languages.
Darpa wants the whole program on a PDA device (I want it on my phone) and wants it ready in 6 months.
That’s some app, but if it works it could take interpreters out of danger and out of the messy equation that the current method of spying creates.