Automated PrOduct GEneration and Enrichment (APOGEE) Navy SBIR FY2012.2 Sol No.: Navy SBIR FY2012.2 Topic No.: N122-136 Topic Title: Automated PrOduct GEneration and Enrichment (APOGEE) Proposal No.: N122-136-0168 Firm: DECISIVE ANALYTICS Corporation 1235 South Clark Street Suite 400 Arlington, Virginia 22202 Contact: Timothy Hawes Phone: (703) 414-5032 Web Site: http://www.dac.us Abstract: Creating information products to answer "Tell Me About" questions requires the ability to identify key pieces of information relevant to a complex set of content requirements. Complicating matters, these key pieces of information are scattered across data stores and buried in huge volumes of data. This results in the current predicament analysts find themselves; information retrieval and management consumes huge amounts of time that could be better spent performing analysis. The persistent growth in data accumulation rates will only increase the amount of time spent on these tasks without a significant advance in automated solutions for information product generation. We propose a system called Automated PrOduct GEneration and Enrichment (APOGEE). APOGEE automates the creation of information products; learning the creation process by example. There are three stages to APGOEE's workflow; first, using clustering and other machine learning techniques APOGEE learns the content models for a range of information products; next, using a search-and-align based methodology, APOGEE maps the content models to the semantic structure underlying unstructured text; finally, APOGEE uses the learned content model and semantic mapping to automatically generate new information products. All this can be done on the fly without requiring predefined information product templates or ontologies. Benefits: The Automated PrOduct GEneration and Enrichment (APOGEE) aims to shift the bulk of effort in information product creation from data discovery and management to analysis by automating the information product creation process. APOGEE uses a three stage process learning content models for information products, mapping content models to semantically enriched text, and using the results of the previous stages to generate concise and complete information products. APOGEE does not rely on predefined, manually created ontologies, markup or templates instead learning to generate information products by example. This means APOGEE has no reliance on manual effort to adapt the system to new scenarios or question types. Additionally, APOGEE uses frame-semantic and thematic models of text to support both content modeling and information extraction yielding result more accurate than competing approaches. Finally, APOGEE uses a filter-by-ranking approach to organize and fuse the information in the resulting product. This approach results in concise information product, presenting enough information to sufficiently fulfill the content model for the generated product but not more. The end result of the APOGEE effort will be a flexible end-to-end system that can be rapidly deployed in new operational settings. Return