Automated Linguistic Analysis Revealing Misrepresentation and Messaging (ALARMM) Navy STTR FY2010.A Sol No.: Navy STTR FY2010.A Topic No.: N10A-T029 Topic Title: Automated Linguistic Analysis Revealing Misrepresentation and Messaging (ALARMM) Proposal No.: N10A-029-0564 Firm: Charles River Analytics Inc. 625 Mount Auburn Street Cambridge, Massachusetts 02138-4555 Contact: Terry Patten Phone: (617) 491-3474 Web Site: www.cra.com Abstract: The proposed Automated Linguistic Analysis Revealing Misrepresentation and Messaging (ALARMM) system significantly advances the state of the art in detecting deception in unstructured data such as Web sites and chat messages. Our research focuses on automated techniques based on linguistic theory that can detect both misrepresentations and hidden messages. The proposed work draws on the field of sociolinguistics for advanced statistical techniques capable of leveraging a wide range of morphological, lexical, and grammatical features. Formal representations from sociolinguistics are used to capture generalizations across types of deceptive language. We propose innovative experiments to provide a quantitative, metrics-based evaluation of the proof-of-concept ALARMM system. Benefits: The proposed research is directly applicable to open-source intelligence analysis as it will detect many types of misrepresentations as well as hidden messages in unstructured text. This research also has several important commercial applications including legal discovery (e-discovery), fraud detection, resume evaluation, and the management of interactive Web sites. Return