Diffeo, A Winning Data Fusion SolutionNovember 9, 2016
Football blaring on the TV. A turkey carcass. Half a pumpkin pie. Sleepy people on the sofa. That’s disparate data the average person can add up to a post-Thanksgiving afternoon.
If you’re the National Geospatial-Intelligence Agency (NGA), disparate data looks like geospatial hazard data, water quality data, data from the World Food Program, FAA images of airports worldwide, wildlife trafficking, Nepal disaster response, U.S. fire data per state, and 20 other data sets.
In this challenge, the NGA sought innovative ways of retrieving and analyzing widely disparate data in different locations, formats, schemas and interfaces. Solutions were rated based on desirability (meeting user needs/desires), expected global utility (usefulness of search results), technical effectiveness (precision and recall, handling special/imprecise data formats) and viability (low cost, easy integration).
Diffeo works like a semi-autonomous research assistant in the familiar environment of the analyst, whether that is email, OneNote, or other data entry tools. Diffeo creates a knowledge graph of people, places, organizations and events, and their relations to one another. The system finds relevant information by fanning out several hops ahead of the user’s current state of knowledge. The system prioritizes content that fills knowledge gaps detected by tracking what the analyst has already put into their in-progress document. The result is machines operating at human-level accuracy, enabling humans to operate at machine scale.
Diffeo uses entity extraction from Rosette text analytics to find people, places, and organizations in many languages. It is one of many startups benefiting from Basis Technology’s Startup Program that provides high quality multilingual text analytics to high-impact, early stage companies.