System and method of using dynamic variance networks

摘要

Systems and methods for determining a location of a target in a document. Information compiled from a training document is created, the information comprising a reference and a reference vector tying each reference to the target. The reference is compared to a new reference in a new document to determine if there are any similar references that are the target, wherein similar references are: position similar, or type similar, or both. When the new reference comprises a typo, an optical character recognition (OCR) mistake, or both, the new reference is still determined to be the target because of the new reference's location.

主权项

1. A method for determining at least one location of at least one target in at least one document, comprising:
creating, utilizing at least one localization module and at least one processor, information compiled from at least one training document, the information comprising at least one reference and at least one reference vector tying each reference to the at least one target, wherein the creating further comprises:
finding, utilizing the at least one localization module, the at least one reference;creating, utilizing the at least one localization module, the at least one reference vector for each reference;performing variance filtering, utilizing the at least one localization module, on the at least one reference and the at least one reference vector from each document to obtain any similar references and any similar reference vectors from all documents; andusing any similar references and any similar reference vectors, utilizing the at least one localization module, to create at least one dynamic variance network (DVN), the at least one DVN comprising at least one reference and at least one reference vector tying each reference to the at least one target; comparing, utilizing the at least one localization module, the at least one reference to at least one new reference in at least one new document to determine if there are any similar references that are the at least one target, wherein similar references are: position similar, or type similar, or both; and wherein when the at least one new reference comprises at least one typo, at least one optical character recognition (OCR) mistake, or both, the at least one new reference is still determined to be the at least one target because of the at least one new reference's location; and applying the information, utilizing the at least one localization module, on at least one new document to determine at least one location of the at least one target on the at least one new document, wherein the applying further comprises:
comparing, utilizing the at least one localization module, any similar references to the at least one new reference on at least one new document to determine if there are any matching references; andusing, utilizing the at least one localization module, any similar reference vectors corresponding to any matching references to determine the at least one target on the at least one new document.