ATL's Knowledge Discovery Capability

I try to keep my posts non-technical, but sometimes I fail. This is one of those times. So it's probably best to skip the article below unless you played Dungeons & Dragons as a child.

Accenture Technology Labs' rather generically-named Knowledge Discovery Capability (KDC) is actually a sophisticated data integration and visualization tool designed primarily for use in research and investigative fields (e.g., science, law enforcement, intelligence, disease management, capital markets, etc).

The KDC stems from a disturbing statistic: Globally, government intelligence analysts spend roughly 80 percent of their time gathering information (from numerous databases) and only 20 percent analyzing it. Further, they have little confidence that the information they find relates to the issues they're investigating. Another, more subtle problem is that current analytical tools answer questions...but they don't generally suggest new questions that might be worth asking. In other words, if you're barking up the wrong tree, today's tools won't tell you you're an idiot.

KDC's pedigree in the Labs began many years ago with the Pocket Exchange, which extracted and integrated information on Accenture personnel from several databases. It continued with the Knowledge Discovery Tool, which scans dozens of databases to enable more effective drug discovery.

So What?

The KDC has a patented front-end that can be deployed on its own or as part of a portal or larger analysis workbench. One application is a prototype the Labs created to show the interplay between a map ("geo-visualization") and a visual network navigator (imagine connecting constellations and then following lines from star to star). The geo-visualization might be used to map disease outbreaks against food shipments against local cargo aircraft flight maps. Putting all of that information on a map simultaneously might answer important questions about the outbreak–but, more than that, the sheer volume of visible data would probably suggest additional questions that might be worth asking.

The network (star) browser is ideal for traversing relationships: from a suspect to flights the suspect has taken to known criminals who were on those flights and so on.

The KDC's back end is a sophisticated data architecture that integrates entities and relationships from a wide variety of sources. Numerous types of connections between people, news items, investigative reports and other entities are inferred, captured and stored for later display either on the map or in the network browser. With the data architecture, what would normally require multiple queries across multiple databases using several different query languages (with some of which your skills may be dismal) becomes a much simpler query and navigation operation that uses just two intuitive metaphors (map and network).

The KDC has non-intelligence applications: Anywhere there's a need to explore complex webs of entities is a potential value-add area. An earlier version of the KDC, for example, is helping the Irish Department of Inland Revenue perform taxpayer audits. (Membership in the Irish chapter of the Accenture Technology Labs Fan Club is modest.)

The KDC is, in short, an interesting tool with a lot of potential. If you'd like to learn more, just get in touch with Accenture Technology Labs' Ryan LaSalle KDC), who is leading the charge in this space.