Federated search engines like Science.gov and WorldwideScience.org reduce the chore of accessing large numbers of heterogeneous databases by simultaneously broadcasting queries to each of the target databases, merging and ranking the results with minimal duplication, and clearly organizing the display with value-added capabilities, such as search result clustering. Time savers as they are, federated search engines typically rely on the basic keyword manipulation retrieval strategies of the target systems in matching queries to documents. Furthermore, federated search engines lack deep semantic knowledge to facilitate exploration and discovery, which is particularly important in searching scientific and technical databases. This SBIR will explore the automatic generation of rich semantic representations of scientific knowledge and the utilization of the resulting knowledge bases to support better quality federated searching, creative exploration and discovery, and question answering in scientific and technical databases. The proposed research will build upon our prior work in utilizing natural language processing (NLP) techniques to construct semantic networks for energy and biomedicine from existing thesauri, such as DOE
