CIA turns to James Bond inspired Q-Tec for gadgets

When the CIA wants cutting-edge technology, it turns to a group of venture capitalists named after James Bond's gadget man - reports The Arizona Republic (USA).

And when In-Q-Tel Inc. wanted a way to sort through mountains of documents quickly and accurately, it turned to an Arizona State University computer scientist to develop software that can recognize and automatically separate handwritten notes, printed material and graphics like maps and logos.

In-Q-Tel - a combination of "intel," short for intelligence, and Q, after the Ian Fleming character from the Bond books - was founded to help the CIA stay ahead of things in the rapidly changing field of information technology.

It is the first time the unusual organization has put money into an Arizona company or Arizona university project, and the ASU project is among just a handful of university programs nationwide that In-Q-Tel has invested in. In-Q-Tel focuses on unclassified technology, but in typical CIA fashion, it is keeping hush-hush about how much money it put in the ASU project and especially about how its parent agency might use the technology.

But one of In-Q-Tel's investment priorities is knowledge management and visualization. Given the massive stacks of documents seized during the Iraq war, it isn't hard to imagine that the ASU technology could help the spy world sort through documents and find the ones that can help intelligence efforts there.

"We're excited that we have technology that is important to national security and the government and has a host of commercial applications," said Peter Slate, chief executive officer of Arizona Technology Enterprises LLC. AzTE is the non-profit corporation responsible for technology transfer, or commercializing professors' inventions, at ASU.

The document-sorting software was developed by Anshuman Razdan and graduate students at the Partnership for Research in Spatial Modeling research center at ASU, often dubbed PRISM. Razdan used his expertise in three-dimensional modeling to go beyond current technology known as optical character recognition, or OCR.

OCR has been around for years as a two-dimensional way to turn printed text into digits that computers can read. The computers can then be programmed to detect keywords and alert humans to documents containing those words.

But OCR works best on typewritten documents uncluttered by hand-scribbled notes, borders and the inevitable coffee stains. And only a fraction of documents fit that bill.

"The fundamental problem is that 80 percent of documents are of mixed quality," Razdan said. "Rarely do you get nice, clean, printed documents."

The ASU technology grew out of PRISM's earlier work categorizing 16th- and 17th-century Spanish manuscripts for ASU's Hispanic Research Center. Because they were dealing with handwriting and a foreign language, PRISM had to find patterns in the three-dimensional curves of the words, then put the patterns into categories.

The research might have stopped there if In-Q-Tel hadn't come to check out what technology AzTE had to offer. It wanted a way to scan documents with a mix of print, handwriting and graphics and to automatically separate out print-only documents so they could be sent through OCR devices, Razdan said.

The In-Q-Tel funding helped pay for the time Razdan and two graduate students spent developing the software, on which a patent is pending.

Last year, the agency sent PRISM several hundred documents that had been scanned in the field. PRISM's software checked for typewritten documents and flagged handwriting, signatures, logos and other elements on the rest.

The "secret sauce," Razdan said, is in looking at documents in 3-D. Printing applies ink to a page at a fairly uniform height, but the human hand deposits ink in such widely varying amounts that the words practically dance on the page, he said. The 3-D process also is not language-specific, making it especially useful for sorting documents in Arabic and other languages where even typewritten characters closely resemble handwriting.

Razdan and his students have gotten the software to the point where it can sort with 95 percent accuracy. Training a computer to read handwriting is down the road, Razdan said.

The technology could someday have commercial uses, too. Libraries could scan their collections and separate handwritten notes and manuscripts from printed materials. Pharmacies could store doctors' handwritten prescriptions in a searchable database, and companies and organizations could create institutional memories from employees' handwritten notes, Razdan said.

AzTE is seeking commercial partners to license the PRISM technology, and it is also considering creating a spinoff company based on it, Slate said.

That is the idea that would work for In-Q-Tel. The agency wanted to tap innovations coming from small companies, and it wanted them to have commercial potential so the costs would ultimately be lower than with custom projects, said Cathy Cotell, vice president of university research.

"We want something that solves the government's problem and Wal-Mart's problem," she said.

The organization identifies technologies, invests in them and arranges for pilot tests within the CIA. If they succeed, In-Q-Tel can negotiate a license for the CIA. Like typical venture capital firms, it takes equity in companies in exchange for its investments and works with other venture capital firms to find the most deals.

One of In-Q-Tel's investments, Keyhole Corp. of Mountain View, Calif., was acquired in October by Google Inc. for an undisclosed amount. Google plans to add Keyhole's digital mapping technology to its Internet search service.

Thanks to `` for the alert.

Discuss this news here...