U.S. defense projects may drive innovations in machine translation
The U.S. Department of Defense's research arm, DARPA, is spending $15 million to create an interpreter robot.
Before you laugh or call your congressional rep about boondoggle projects, think of other once-seemingly wild ideas that first came from the U.S. military—the Internet for one (sorry Al Gore) and then there's the DARPA "PAL", the Personal Assistant that Learns, which eventually evolved into the iPhone's silken-toned Siri.
It's no mystery why the U.S. Department of Defense is interested in machine translation. Americans tend to be monolingual and so is the country's military. And where the U.S. fights, English usually isn't the dominant language.
Enter technology. After trying hand-held translation tools, DARPA kicked it up a notch with the BOLT project (Broad Operational Language Technology). According to the Defense Industry Daily
, the project's goals include creating technology that cannot only translate multiple languages but also enable bilingual communication through text and speech—person to machine (like Siri does) and person to person.
Yes, that means an interpreter robot. As Wired
put it, the Defense Department wants to build a real life C-3PO. And there are other translation projects in the works, such as the recently announced $5.9 million contract with Raytheon BBN Technologies
, to create a real-time English translation of documents, including handwritten notes or images with text on them.
What does this all mean for the translation industry? It's probably a Star Wars fantasy that there would be interpreter robots to accompany international travelers, at least anytime soon, but there are other results that might not be so far-fetched.
At the very least, the Defense Department's focus on Arabic and Mandarin should improve machine translation in those languages. If successful, the Raytheon project would make dealing with a range of documents easier. And beyond the robot, the BOLT project also involves quite a bit of document translation, with a goal of reaching 90% accuracy. And it’s not so crazy to think BOLT's voice-translation technology could evolve into a consumer application as Siri did.
Still it's hard to ignore BOLT's "Activity E"— research into deep semantic language acquisition, with the intent to give the robot the ability to assess what it sees and touches as well as analyze actions and consequences—in other words, to think.
Kind of grabs the imagination, doesn't it?