Sorting app helps make sense of worldwide intelligence data
- By Vandana Sinha
- Apr 22, 2003
The CIA's In-Q-Tel venture capital arm has handed $500,000 to a California software company to add more foreign languages to its mohoClassifier Web-based filtering system.
The CIA uses the classification software from Mohomine Inc. of San Diego to sift patterns in e-mail and files arriving via the Internet, intranets and extranets. The software gradually learns to ignore junk leads and highlight priority leads'in some cases routing them to the right people and copying key points from one file to another for easier analysis.
MohoClassifier sifts French, Italian, Spanish, Portuguese and German, as well as a few other languages that do not use the Roman alphabet. In-Q-Tel wants additional unspecified languages.
'I would imagine they're looking at Arabic and Middle Eastern languages,' said Laura Ramos, director of enterprise portal research at Giga Information Group Inc. of Cambridge, Mass.
Regardless of language, the filtering software detects patterns in character configurations and assigns weights according to priority. It learns by example, applying algorithms to follow the lead of up to 20 past sorting and routing choices that were manually approved.Spam scanner
Organizing 100M worth of data previously took a CIA analyst hundreds of hours. The software can scrutinize nearly 10,000 e-mails per minute.
For the CIA, which Ramos called 'an organization with one of the largest unstructured data problems imaginable,' the time savings is significant.
'Much more information comes in than any human being can handle,' said Kim Cook, director of technology assessment for In-Q-Tel, which has invested $1.5 million in Mohomine since early last year.
Cook would not divulge details about the technology's use at the CIA, but she said the automated classifier helps analysts focus on search results rather than on the search itself.
'Instead of spending their time alphabetizing files,' she said, 'they can spend time really analyzing the material in there.'
In-Q-Tel has funded companies with similar products, including Stratify Inc. of Mountain View, Calif., and Intelliseek Inc. of Cincinnati. But Cook said Mohomine's processing speed, back-end support and a claimed sorting accuracy of close to 98 percent made it a candidate for an additional contract.
A new Web platform has expanded Mohomine's classification universe beyond its Microsoft Windows 2000 operating system. MohoClassifier 2.3 costs $50,000 to $1 million depending on the size of a project.