DARPA Launches Translation Software Initiative

Slideshow: Next Generation Defense Technologies (click for larger image and for full slideshow)

The research arm of the Department of Defense (DoD) is working on new technology that will allow U.S. military to translate face-to-face conversations in real time, as well as accurately decipher intelligence and communications data in voice, video, and print.

The Defense Advanced Research Project Agency (DARPA) is seeking proposals for its Broad Operational Language Translation (BOLT) program, which aims to build software specifically to translate various dialects of Arabic as well as Mandarin Chinese.

The agency hopes to tackle not only formal aspects of the languages but also informal and colloquial ways of speaking, such as what is found in emails or occurs in conversational speech, it said.

In addition to translation of various forms of media, BOLT also will develop machine translation that people can use face-to-face and in real time to engage in complex conversations with Arabic and Mandarin Chinese speakers that don't speak English, and vice versa, according to the agency.

This aspect of BOLT aims to evolve a translation program DARPA already has called the Spoken Language Communication and Translation System for Tactical Use (TRANSTAC), which allows for spontaneous two-way communication by translating short phrases.

Other program goals include the capability to retrieve information from a variety of multilingual sources using natural-language English queries, and research both deep semantic language acquisition and basic technologies such as parsing and language modeling, according to a Broad Agency Announcement posted on FedBizOpps.gov.

In addition to TRANSTAC, DARPA has several other translation projects in progress. One is the Global Autonomous Language Exploitation (GALE) program, which has developed software to translate formal news in Arabic even more accurately than human translators.

Another program, Robust Automatic Transcription of Speech (RATS), is aimed at building a system that can accurately determine speech activity, identify a speaker by his or her voice, and pick out keywords even in cases where speech is garbled or where there is noisy interference.

Still another program, called the Multilingual Automatic Document Classification Analysis and Translation (MADCAT), is working on technology to automatically convert foreign language text images into English transcripts.

BOLT has three technical areas, and DARPA is seeking proposals on all of them. The first technical area of the program is to develop algorithms and integrated systems to support the translation, and the second will focus on data collection. The final technical area is evaluation of the solution. Each area is divided into three phases that will last about 12 months each.

DARPA will make multiple awards for each technical area to support BOLT, and not all contractors will work on every area of the project, it said. The final deadline for proposals is Aug. 2.