Address Locator
This locator method finds addresses on structured, unstructured, and semi-structured machine-printed US documents. The individual elements of the address blocks are usually located close together and can span across a single horizontal line, or span two to seven lines forming a vertical block. This locator finds these blocks on all pages of a document and returns them as a list of alternatives. The alternatives include subfields for each individual piece of extracted information, such as the name, street, or city.
The list of alternatives is sorted by confidence. If part of the address block is illegible, the confidence for that alternative is lower because of the missing data.
If multiple addresses are found on a document, it is up to the solution integrator to build in some functionality that can narrow the choice of addresses. For example, they can set up regions to limit the search area or apply a simple script.