Entity resolution helps FBI's N-Dex system connect the dots
- By David Essex
- Oct 12, 2009
How can police use computers, which are famously literal and unforgiving, to track down James Smith if some of his records say Jim Smith? The answer is entity resolution, a software process that can discern the likelihood that James and Jim are the same person. For the National Data Exchange, the FBI turned to Initiate Entity Resolution from Initiate Systems.
Raytheon and FBI sources agree it’s the software that makes N-Dex special, allowing intelligence and law enforcement agencies to connect the dots among pieces of information spread out across disparate sources. “Agencies can link together information that had been thought to be unrelated, providing the potential to identify suspected persons of interest, link events to persons or identities or connect already known persons of interest to others not yet identified,” said Scott Schumacher, senior vice president and chief scientist at Initiate Systems.
“Entity resolution is the process of determining when disparate pieces of information, usually records from different sources, refer to the same object, such as a person, organization, location or event,” Schumacher said. “Identity resolution is entity resolution applied to persons. In N-Dex, it establishes associations between multiple incidents — for example, a witness for an incident in one county may have been a suspect in an incident in another jurisdiction. This allows investigators to tie together related incidents and share related information.”
Schumacher said the software applies statistical decision theory to tell whether two records refer to the same entity and employs pattern matching and a “data-adaptive” framework to adjust to each problem. But the real trick of the software might be its ability to identify and manage relationships across different types of information. Unlike system-centric approaches that manage information siloed by data type, Schumacher said Initiate Entity Resolution takes an entity-centric approach. Speed, accuracy, flexibility and scalability are improved by calculating likelihoods instead of applying rules, he said.
Raytheon program manager Andre Haynes said his company brought in Initiate engineers to help develop algorithms that would work with N-Dex data, and to tune them for performance. N-Dex development manager Jim Preaskorn called it one of the team’s biggest challenges. “Police officers like to abbreviate things,” Preaskorn said, so the search algorithms had to be refined to resolve their meanings.
-- David Essex
David Essex is a freelance technology writer based in Peterborough, N.H.