US Patent 9372916 Document template auto discovery

Methods and apparatus for generating a template for automatic data capture are described. The method comprises determining locations of a plurality of data fields in a first document, wherein the plurality of data fields are identified based, at least in part, on structured data associated with the first document, identifying at least one second document that includes the plurality of data fields in locations similar to those determined for the first document to produce a set of documents, determining locations of a plurality of anchorboxes describing common text elements of the set of documents, and generating the template, wherein the template describes locations of the plurality of anchorboxes and locations of the plurality of data fields.