IARPA Babel program seeks to provide translation and intelligence analysis for any known language

WASHINGTON, 10 April 2011. U.S. intelligence analysts want to know what everyone is talking about.Officials of the U.S. Intelligence Advanced Research Projects Activity (IARPA) in Washington are asking companies and universities for help in a project to transcribe and translate any language quickly for further analysis.The IARPA Incisive Analysis Office issued a broad agency announcement (IARPA-BAA-11-02) last week for the Babel program, which seeks to develop speech-recognition technology for any human language that will help U.S. intelligence analysts search and process massive amounts of recorded speech.Babel is intended to develop new technology able to transcribe any new language within one week for keyword search of massive amounts of speech recorded in challenging real-world situations, IARPA officials say.

Today's speech-transcription systems for intelligence analysis primarily target conversations in English, and have relatively poor performance with other languages, IARPA officials explain. These systems, furthermore, often have taken years to develop and cover only a few languages of the world. The IARPA Babel program seeks to change all that.

Although speech researchers for decades have assumed that the techniques they use for analyzing English are as effective for most other languages, current research suggests that other approaches are necessary, IARPA officials say.

In addition, an increasing number of languages worldwide are of critical importance to intelligence analysts, and for many of these languages there is no existing speech technology with which to extract information vital to U.S. and allied national interests. IARPA experts are trying to create keyword search technology for many languages to enhance the intelligence analysts's ability to make sense of that data.

The Babel program's chief goal develop speech recognition capability quickly for keyword search in a previously unstudied language, working with speech recorded in a variety of conditions with limited amounts of transcription.

Companies chosen to participate in the Babel program will receive government-furnished speech data to help with research. Babel will address a broad set of languages from families such as Afro-Asiatic, Niger-Congo, Sino-Tibetan, Austronesian, Dravidian, and Altaic.

Key technical challenges revolve around methods that are effective across several different languages, speech recorded in noise, limited amounts of transcription, and keyword search algorithms for speech.

Several different contractors are likely to be chosen for the Babel program. Companies and universities interested should respond no later than 7 June 2011 to ODNI/IARPA, Attn: Dr. Mary P. Harper, Gate 5, 1000 Colonial Farm Road, McLean, VA 22101.

For questions or concerns contact Dr. Mary P. Harper by phone at 301-851-7673, or by e-mail at [email protected]. More information is online at https://www.fbo.gov/notices/5e8995ba03eac9b699fca5d24fb78590.