Language Technology

Bonnie Dorr (left), program manager in DARPA’s Information Innovation Office (I2O), shakes hands with Henry Kautz, past president of the Association for the Advancement of Artificial Intelligence (AAAI), upon her recent induction as an AAAI Fellow. Each year, AAAI bestows the lifetime honor of Fellow on only a handful of researchers for their exceptional leadership, research and service contributions to the field of artificial intelligence.

Understanding local languages is essential for effective situational awareness in military operations, and particularly in humanitarian assistance and disaster relief efforts that require immediate and close coordination with local communities. With more than 7,000 languages spoken worldwide, however, the U.S. military frequently encounters languages for which translators are rare and no automated translation capabilities exist. DARPA’s Low Resource Languages for Emergent Incidents (LORELEI) program aims to change this state of affairs by providing real-time essential information in any language to support emergent missions such as humanitarian assistance/disaster relief, peacekeeping and infectious disease response. The program recently awarded Phase 1 contracts to 13 organizations.

Expanded global access to diverse means of communication is resulting in more information being produced in more languages more quickly than ever before. The volume of information encountered by DoD, the speed at which it arrives, and the diversity of languages and media through which it is communicated make identifying and acting on relevant information a serious challenge. At the same time, there is a need to communicate with non-English-speaking local populations of foreign countries, but it is at present costly and difficult for DoD to do so.

Department of Defense (DoD) operators and analysts collect and process copious amounts of data from a wide range of sources to create and assess plans and execute missions. However, depending on context, much of the information that could support DoD missions may be implicit rather than explicitly expressed. Having the capability to automatically extract operationally relevant information that is only referenced indirectly would greatly assist analysts in efficiently processing data.

The U.S. Government operates globally and frequently encounters so-called “low-resource” languages for which no automated human language technology capability exists. Historically, development of technology for automated exploitation of foreign language materials has required protracted effort and a large data investment. Current methods can require multiple years and tens of millions of dollars per language—mostly to construct translated or transcribed corpora.