Notice ID ECON-21-0041
“The U.S. Census Bureau is seeking to learn more about vendor capabilities and core competencies as it relates to performing Optical Character Recognition (OCR) to capture handwritten names and additional handwritten information from digital images. In the event of a requirement, the vendor will be expected to be a subject matter expert in OCR of handwriting, and especially the use of OCR for capturing handwriting historical images obtained from scanned microfilm. Given the complexity of capturing handwriting and the volume of data needed to be captured (e.g. more than 850,000,000 names and responses to capture), the interested vendor would need to have the experience and ability to provide complete data. The vendor must be able to work effectively with the collaborative team of Census Bureau researchers to develop the OCR and apply it within the Census Bureau’s computing environment. It is envisioned that this requirement will be for a base period of 12 months, plus options for an additional two twelve-month period of performance.”
“An interested vendor would be expected to develop and execute OCR for example, with 1960-1990 Decennial Census images as indicated in the service activities listed below:
- Develop OCR for each Decennial Census. The contractor will develop specific OCR methods for each Decennial Census using training data and training images, along with name dictionaries and other metadata. The contractor will also develop the metadata as needed for the OCR.
- Perform OCR on test data. The contractor will test OCR methods on a set of test images in order to assess the quality of OCR throughout development. The output will be delivered to the DCDL Project Manager for analysis. Multiple OCR tests are expected for each decennial census, and further development of methods will take place between testing.
- Work with Census Bureau staff on refining OCR methods. The contractor will work closely with DCDL project staff throughout the development of OCR. Analytical review, quality control, and quality assurance systems will be developed along with Census Bureau staff. The contractor will respond within 24 hours to all requests for information/assistance from project staff. The contractor will also participate in weekly status update meetings so that the contractor can provide subject-matter expertise and answer questions in a group setting.
- Produce preliminary captured data files. The contractors will deliver preliminary captured data to the DCDL Project Manager for the 1970 and 1980 records by the end of year 1 and for 1990 and 1960 by the end of year two.
- Produce Final captured data files. The contractor will perform OCR on the full corpus of 1960, 1970, 1980, and 1990 captured images. The resulting data will be delivered to the project team by the end of year three.
- Document and present OCR methods and outcomes. The contractor will document the OCR methodology and report it to project staff as requested. At the end of the project, written documentation of the processes used to capture the data will be created and given to project staff. The contractor is expected to develop academic papers on the project in collaboration with Census Bureau staff. All publications will require full Census Bureau disclosure and policy review. Presentations to Census Bureau staff are expected, along with presentations at academic conferences.”