In this example, an analyst has assembled a collection of approximately 3500 intelligence messages, of four different types, describing a (simulated) situation in Bosnia-Herzegovina. The four message types include human intelligence reports (IIRs), tactical intelligence reports (TACREPs) describing voice radio communications between military units, reports describing movements of military vehicles (REXREPs), and reports describing radar activity (TACELINTs). As shown above, the material has been organized into two separate Starlight databases, or InfoSets, one containing the REXREPs and TACELINTs (presented as a Starlight Discrete View), and the other containing the TACREPs and IIRs (presented as a Concept View). The analyst has also assembled supporting image and map data to provide a visual context for certain aspects of the information under study. The analyst's challenge is to reassemble these elements into a coherent picture of activity in the region, both to characterize the current situation and for predictive purposes.
Sounds like an awesome program, if only it could be used a lot more widely in our intel agencies! I know that Rep. Weldon says in his book he has been agitating for extensive intel data-mining and 'fusion' programs across all gov't classified databases since 1997-98....he says there are at least 32 major classified databases which are only accessible to their own internal searches. Of course, security and privacy are always big issues, and spy-moles like Aldrich Ames and Robert Hanssen made the various agencies guard their data even more tightly, I'd suspect.... AD only focused upon UNclassified data, as I understand it, but one of Weldon's main points is that data-mining needs to be done across all available classified and unclassified data so that all possible bits of info can be pulled togther.... e.g., someone like ATTA might have left tidbits of data in a dozen or more different databases, some classified and others unclassified..... [are Starlight and similar programs being used much at all for intel analysis and counter-terrorism since the demise of AD??]
One thing I worry about is the fact that names from Arabic, Persian, and a variety of other languages may be translated into English in a variety of ways, I believe. Is there software that can recognize numerous variants of the same name, and especially with all the abu-terrorist, al-bozo, el-scumbag, etc. variations that exist..... can intel agencies really search adequately, and are "watch lists" able to deal with such variations and also mis-spellings???