Key to achieving high recognition rates is understanding address structure.
As OCR has become a commodity, the key to successful address interpretation has become less about recognizing characters and more understanding the information contained within the address structures, so that the pool of addresses that are able to be ‘understood’ is as large as possible.
Using address element analysis and categorization, the pool of valid address structures can be expanded from a few dozen to several hundred. This means that more addresses can be interpreted and sorted to the deepest level possible.
PASS will work with the postal authority to perform tokenization and address element analysis using MIDAS, a web based truthing and analysis tool.
Furthermore, the results of truthing and address element analysis create the basis for an ongoing continuous improvement platform.