Intelligence Question Answering System (INQA)

Client: Intelligence Advanced Project Research Activity (IARPA), via Naval Research Laboratory, Interactive Systems Section (Code 5512)


Objective: Develop a demonstration system for showcasing and evaluating the advances in next generation of question answering technologies to the office of the Director of National Intelligence (ODNI).  Design and conduct human subject studies to demonstrate the benefit of such technologies over a baseline keyword search technology.


Our Approach: Knexus team lead by Dr. Kalyan Gupta along with System Architect Mr. Philip Moore supported the Government Evaluation Team with architecting and designing a federated system called Intelligent Question Answering (INQA). INQA provided a seamless integration architecture for five research teams developing advanced natural language based question answering system capable of handling cross language queries. That is, for example, search Arabic documents and return an English answer to a question asked in English.

Using a service oriented approach, the Knexus development team designed a browser based Q-A system for use and evaluation by intelligence analysts.  The Knexus team successfully  coordinated the development of component services and their integration into the architecture in a timely manner, and within budget. They supported  the Government evaluation team to design and conduct human subject studies. They delivered their conclusions and their recommendations in a report to the program.  


Solution We Delivered: The proof-of-concept prototype was deemed highly successful in demonstrating a future concept of operations for the Intelligence Analysts. It was also a key component of a thorough and fair  scientific evaluation. It showed how such advanced technologies could effectively reduce effort and support analysts with preparation of intelligence reports.   The success of the system resulted in its selection and subsequent demonstration for the US Congress  at  the Tech Expo, at  Capitol Hill, in 2008. 


Key Concepts:  Federated Decision Support,  Human Subject Study, Cross Language Retrieval, Intelligence Analysis


Client Feedback: “I wanted to take this opportunity to tell you that I am more than pleased with employees and work provided by your company, Each and every individual is performing at levels that exceed my expectations… not only are they excellent researchers and scientists, but they also clearly provide team support for each other” Dr. Dennis Perzanowski, Section Head, NRL PI for AQUAINT Evaluation Team.