Census releases draft RFC for DCS

The Census Bureau is one step closer to awarding a contract for the decennial census image-based data capture system with the recent release of a draft request for comments.

The draft RFC, issued this month, overhauls an earlier document the bureau released last November for vendor comment. As a result of those comments, changes have been made to the scope of the contract, how the procurement will be conducted and the system requirements.

The Data Capture System 2000 will use digital electronic imaging for the first time to process around 1 billion pages of census forms in 100 days for the 2000 census. What also makes DCS 2000 challenging is that it must be fully operational and able to process the required workload from the first day of data capture.

The system will process incoming census forms, digitally capture and process form images, automatically convert the image data to text-based data and edit data that cannot be automatically converted. DCS 2000 will contain five subsystems: check-in, workflow, imaging, optical recognition and keying from image.

Forms will be scanned and then passed through optical mark recognition and optical character recognition engines to convert the images data into ASCII. When the forms cannot be scanned, or where OMR/OCR engines cannot interpret the data on the forms, it must be manually keyed in. When the data is captured, the files will be transmitted to the bureau headquarters for post-data capture processing. The bureau will then merge data capture files.

According to the draft RFC, the DCS 2000 procurement will now be conducted in two phases. The first will cover the design and development of a prototype DCS 2000 system, and the second will include a follow-on contract for full-scale production development of the prototype. Phase two may also include the acquisition and operation of one or more data capture centers.

The bureau estimates it could have about four centers throughout the country, each of which would house a DCS 2000. It expects to award the fixed-price contract to one or more vendors.

DCS 2000 is a first for the bureau in several ways, according to Arnold Jackson, information resources management director at Census.

"We never used digital electronic imaging before, and we're going for an industry solution rather than in-house," Jackson said.

The bureau is considering leasing DCS equipment because the system will essentially be used only for the 2000 census. It also expects to award several more contracts for technology to support other 2000 census activities, Jackson said.

There are many benefits of using digital electronic imaging, said Robert Marx, associate director for the decennial census. "There is a big cost benefit," Marx said. "We'll be able to do the job with less money and fewer people."

Imaging "will do all the [optical] mark sensing and read a fairly good proportion of the hand-written entries and convert to ASCII," Marx said. The previous system captured images on microfilm and could not read handwriting.

The DCS 2000 will also provide "heads up" keying, in which only the machine-unreadable handwritten part of the form appears on a computer screen so the correct information can be keyed in quickly.

"We don't have to move paper around to keyers, and they don't have to turn pages," Marx said. He added that technology has advanced to the point that the bureau will be able to buy a "nearly off-the-shelf" system rather than develop it in-house.

Vendors seem excited about the contract, which could be worth $50 million to $100 million.

Jim Everett, vice president of government sales for ScanOptics, said a multiple-award design phase is a good idea because it reduces some of the risk for the whole program. He also likes the fact that "the procurement activity is closer to what the business would like to see as a more definable time table."

Census expects to issue a request for proposals in June and make an award in September of this year. The award was originally scheduled for next year.

The transition to phase two will take place around October of next year. Data capture centers will start to open around November 1998.