Case Study

Counting Heads for Uncle Sam — U.S. Census conducted utilizing recognition software of Open Text Document Technologies

The Census Bureau is responsible for collecting and making available timely, relevant, and quality data on the population and economy of the United States. Every ten years, the Census Bureau of the U.S. Commerce Department conducts a nationwide survey of the country's population. The last such census, the 22nd in U.S. history, was held three years ago. Open Text Document Technologies participated in the survey, providing its RecoStar software to scan the 140 million questionnaires.

The "Year 2000 Decennial Census" project was the most comprehensive census in the history of the U.S. Four processing centers (Baltimore, Pomona, Phoenix and Jeffersonville) scanned a total of 140 million multi-page forms over a period of 171 days. At peak times, more than six million forms were delivered to the centers every day. The U.S. Commerce Department required that all the incoming forms be registered with the help of barcodes within two days. The aim here was to quickly find out which households had not yet been reached, so that follow-up measures could be implemented immediately. Afterwards, the 85 percent of the forms given the highest priority (registered according to last name) had to be scanned within three months. The fully evaluated data then had to be presented to the President by December 31, 2000. To achieve optimal statistical results, at least 98 percent of the handwritten fields had be recognized without any error.

Open Text Document Technologies provided approximately 200 RecoStar software licenses for recognition of the handwritten entries on the forms. Comparisons with dictionaries helped optimize results. A particular challenge was the large number of documents that had to be scanned and the need to have the software function smoothly with other system components. The recognition software was able to process up to 83 percent of the fields on the handwritten forms with a very low error rate of only 0.58 percent. The error-free recognition rate of over 99.4 percent of all data significantly exceeded the demands of the Census Bureau, coming in at less than 50 percent of the maximum allowable error rate.

Open Text Document Technologies(then Captaris Document Technologies) was named "Supplier of the Year" by the project's general contractor, Lockheed Martin Mission Systems. As a result of its successful participation in America's "Year 2000 Decennial Census," Open Text Document Technologies has now been awarded further contracts for census projects. The company's experience will thus be applied to benefit other countries as well. Open Text Document Technologies' RecoStar character recognition software is currently used in, among other places, the UK, Austria, Australia, Brazil, Croatia, Spain, Kazakhstan, Switzerland, Slovakia, and the Czech Republic.

All company and product names are registered trademarks.

Related Documents

Case Study "US Census" (English - PDF)
Case Study "US Census" (Deutsch - PDF)