download pdf version

Introduction The idea that humans will be able to interact with computers using conversational language has long been a favorite subject of science fiction. Hal 9000 (2001: A Space Odyssey), C-3PO (Star Wars), and the Enterprise computer (Star Trek) all communicated with computer precision while understanding and using the subtleties of spoken language. This reflects the persistent belief that spoken dialogue is the most natural and powerful user interface to use with computers. There remain, however, significant technical challenges to be solved before we approach the human-computer conversational capabilities possessed by C-3P0. Challenges The challenges of creating conversational computer systems are as great as understanding both the complexity and subtility of natural spoken languages. Spoken language understanding engineers still face many challenges, including: (1) speech recognition (what the person said), (2) speech understanding (what the person meant in context), and (3) dialogue management (what is the appropriate response). ATL and Spoken Language Understanding Lockheed Martin Advanced Technology Laboratories (ATL) has more than a decade of spoken language understanding research experience, developing systems that have made significant progress solving these challenges. Included in this experience are nearly a dozen current or successfully completed programs for the Department of Defense and other Federal agencies. WIRE The Wearable Intelligent Reporting Environment (WIRE)—the first wearable, self-contained spoken language understanding system — is an excellent, functional example of ATL's spoken language understanding technology. Wire allows the warfighter to complete a field intelligence report as if he were talking to another person-using national conversation consistent with the situation rather than having to memorize commands and prompts. WIRE understands the context of a statement. When the warfighter says "I just saw five armed insurgents," pauses and then says, "They were traveling east," WIRE understands who "they" are and associates the direction of movement with the insurgents. The system will also ask the warfighter for any additional information that is needed to complete the report. The warfighter can mention observations in any order. The system assembles the observations into a standardized report format to be used by the command post to quickly assess the situation and respond if necessary—or plan future actions in a timely manner. Other future applications for ATL's spoken language understanding expertise include: Cockpit workload reduction in combat aircraft. Field casualty/medical reporting/logistics. Directed analysis of multisource intelligence. Law enforcement reports and database access.