The goal of the Automated Counterdrug Database Update (ACDBU) effort is to streamline and automate the counterdrug and maritime tracking operations at HQ US SOUTHCOM by providing an integrated system that automatically updates temporal and spatial displays with events extracted from live intelligence message traffic. The ACDBU system provides interfaces to a variety of message sources and is capable of extracting data from all United States Military Text Format (USMTF) formatted messages as well as ordinary plain text (i.e., unformatted) messages. The system utilizes state of the art techniques in natural language understanding (NLU) to extract data from plain text messages. All extracted data is automatically formatted and inserted into the appropriate counterdrug databases.
The initial ACDBU system was installed at HQ US SOUTHCOM, Miami, Florida, in November of 1998 and wwas used to support missions at HQ and the (then-) Joint InterAgency Task Force (JIATF) South in Panama. On-site support is provided for the system at HQ and incremental updates were made to the system throughout 1998 and 1999.
Under the ACDBU effort, the following systems are being integrated to provide US SOUTHCOM with an automated message handling and database update capability: Defense Intelligence Threat Data System (DITDS); Generic Intelligence Processor (GIP); Identifinder; and Timeline Analysis System (TAS).
For this effort, the DITDS subsystem will be configured to collect LOCATOR, TACREP, and SANDKEY messages output by Communication Support Processor (CSP) and place them in a message queue for subsequent processing by the GIP subsystem. LOCATOR and TACREP are formatted USMTF messages containing mostly spot reports of various vessels. SANDKEY messages are unformatted, plain text messages, and generally contain a variety of information.
The GIP subsystem is the heart of the ACDBU system. It provides the interfaces to each of the other ACDBU subsystems. It performs the data extraction and validation of all formatted USMTF messages, provides the interface to Identifinder subsystem for the extraction of data from free text messages, generates the appropriate TAS event records from the extracted data, and delivers the records to the TAS subsystem. The GIP subsystem is also highly reconfigurable. Non-expert users can quickly and easily adapt the message processing characteristics of a GIP pipeline to handle new requirements as they surface.
GIP sends any free text data to the Identifinder subsystem which extracts specific categories of information such as names, organizations, locations, dates, and monetary values. The extracted data is sent back to GIP where it is translated, formatted, and sent to TAS for display to the user. Under this effort, Identifinder's finite-state, rule-based technology is being replaced with the latest learning algorithm techniques (i.e., learn by training on sample text with the correctly annotated answers) which will provide the system with language independence, support for multiple text modalities (upper case, lower case, speech format), support for an unlimited number of data categories, and simpler maintenance.
The analysis component of ACDBU is TAS which graphically depicts events that have been extracted by the GIP subsystem on either a timeline or map. TAS operators use the maps and timeline displays to quickly access the current situation in particular areas and to do predicative analysis and mission planning. TAS operators can optionally invoke the GIP/TAS Message Utility (GTMU) to view the original message and generate additional TAS events as necessary. The GTMU aids the user in creating additional TAS events by color coding any Identifinder extracted data (names, organizations, locations, etc.) that appears in the original message. Any additional TAS events that are created within GTMU are sent to GIP for proper formatting and delivery to TAS.
The ACDBU system provides US SOUTHCOM with a real-time, automated message processing and database update capability that will significantly reduce the amount time that is currently spent manually reviewing and entering data into the various counterdrug databases.