DARPA Computational Cultural Understanding (CCU) Program

The Department of Defense (DoD) operates globally and is in constant contact with diverse cultures, and cross-cultural miscommunication can derail negotiations, incite hostile discourse, or even lead to war. Despite advances in machine learning and multimedia analysis, existing human language technology is inadequate for cross-cultural communicative analysis.

DARPA wants AI-enabled systems to go beyond simple translation, leveraging deep social and cultural understanding to assist communication. The aim is to achieve relative parity with human interpreters, who provide cultural insight in the translation process, ultimately assisting negotiators and analysts in the field.

The program is organized into two technical areas:

• TA1 (Sociocultural Analysis): Focuses on sociocultural norms discovery, cross-cultural emotion recognition, and detection of impactful changes in sociocultural norms and emotions.
• TA2 (Cross-Cultural Dialogue Assistance): A framework for a sociocultural dialogue assistant to help monolingual operators have successful interactions in cross-cultural settings.

Please review the CCU Broad Agency Announcement (BAA) for more detailed information on the program.

Evaluation Plans

NIST is tasked to evaluate the performance of systems developed for CCU. The program is divided into two 18-month phases with six evaluation cycles total. The evaluation specification document  describing the evaluation protocol for each evaluation cycle is listed below:

• MiniEval1/Eval1
• MiniEval2
• Eval2
• MiniEval3
• Eval3

Contact

Please email nist_ccu [at] nist.gov (nist_ccu[at]nist[dot]gov) for any questions or comments regarding the evaluations for the CCU program.