Scientific Computing Seminar

The second part is on question answering in the Joint Battlespace Infosphere, or from sensor networks. The Government?s question answering vision and roadmap documents describe a five year program for research and development for question answering systems focusing on supporting the needs of intelligence analysts. The mission of DARPA?s Office of Information Exploitation (IXO) program is to ?. . . develop sensor and information systems with application to battle space awareness, targeting, command and control, and the supporting infrastructure required to address land-based threats in a dynamic, closed-loop process.? IXO is developing 1-, 5-, and 20-year vision statements to meet the challenges of these systems. These dynamic information environments require intelligent middleware to broker services to connect information users and sources. Users pose natural language questions, which must be translated into the query languages and ontologies of the heterogeneous systems making up the JBI. While technologies in this area will build on current DARPA programs providing tools for efficient human creation of ontologies, because of the dynamic, rapidly changing environment represented by the JBI, it is necessary that more automated approaches to semantic interoperability be developed, as well. While much work on semantic interoperability, scalability, and query optimization in such environments has come from the database community, it is argued that approaches based on probabilistic information retrieval and on computational linguistics can play a role, as well.