Sign In

Communications of the ACM

ACM TechNews

How DARPA Plans to Decrypt the Languages That Computers Still Don't ­nderstand

View as: Print Mobile App Share:
A computer with a question.

The Low Resource Languages for Emergent Incidents project is aimed at building a hardware/software platform that can crunch through vast volumes of unknown speech and text in order to extract real meaning.


The U.S. Defense Advanced Research Projects Agency's (DARPA) Low Resource Languages for Emergent Incidents (LORELEI) project, which will officially launch in May, is an effort to aid interpreters by decrypting speech.

The aim of the project is to construct a hardware/software platform that crunches through vast volumes of unknown speech and text to automatically extract real meaning.

LORELEI program manager Boyan Onyshkevych says DARPA wants the system to be capable of producing basic results as soon as a day after exposure to a new language.

The Institute for Human and Machine Cognition's Bonnie Dorr sees a two-pronged challenge in LORELEI's low-resource analysis, in which a dataset needs to be understood as well as built. "In [low resource] speech, you have no idea what's coming your way," she says. "If you find documents, you have no idea what the nature of those documents are."

Not only must data be collected, it also must be associated with as much metadata as possible. DARPA plans to share LORELEI with the general public so it can collect data while concurrently helping in dire situations across the globe, such as assisting U.S. troops in Afghanistan and other war-torn regions where interpreters faces many barriers to local diplomacy and intelligence gathering.

From Motherboard
View Full Article


Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA


No entries found