Back to results
US 2022/0067828 A1Published application
Validation method and system to improve data accuracy
AssigneeSoftWorks Al, LLC
InventorsAri Gross, Matthew Joshua Khan Persad, Yunhao Shi, Perry Kangoun, Talya Klein
Life of this patent
Estimating the term…
Abstract
An automated method and system for validating (cross-validating) data fields in an electronic document, such as a document that has been passed through an optical character recognition (“OCR”) or Intelligent Document Recognition (“IDR”) system or software, to improve accuracy of the electronic document.
Claims26
An automated document-analysis method comprising the steps of:
- automatically selecting a data field in an electronic document;
- automatically identifying each occurrence of the selected data field in the electronic document;
- for each occurrence of the selected data field in the electronic document automatically obtaining a data value and a confidence value;
- grouping the identified occurrences of the selected data field into a set of data-field groups based on their respective data values;
- for each data-field group in the set of data-field groups, determining a number of occurrences of the selected data field and the data-field group's average confidence value;
- applying a first set of criteria to each data-field group in the set of data-field groups, the first set of criteria comprising at least one of
- (i) a number of group-members, and
- (ii) a confidence value threshold;
- based on a result of the applying step, designating one data-field group in the set of data-field groups as a global-value group, increasing the average confidence value of the global-value group;
- for an at least one occurrence of the data field in the global-value group, increase the data field's confidence value;
- obtaining a set of modification criteria;
- applying a set of modification criteria to each occurrence of a data field in a first data-field group, other than global-value group, in the set of data-field groups;
- and
- depending on a result of applying the set of modification criteria, changing an at least one of the data value and confidence value of an at least one data field in the first data-field group to a corresponding value of a data field in the global-value group.
11 more elements
An automated document-analysis method comprising the steps of:
- automatically selecting a data field in an electronic document;
- automatically obtaining a data value and a confidence value of the data field;
- automatically identifying an at least one business rule applicable to the data field;
- automatically applying the at least one business rule to the data value;
- automatically determining a confidence level for the data value based on a result of the business rule applying step;
- and
- adjusting the confidence value of the data field based on the determined confidence level.
4 more elements
An automated document-analysis method comprising the steps of:
- receiving an electronic document comprising a plurality of sections;
- classifying the plurality of sections;
- for each classified section, applying an at least one of (i) an intra-class rule;
- and (ii) an inter-class rule;
- wherein the intra-class rule comprises an at least one of (i) a single-subclass versioning rule;
- and (ii) a date versioning rule;
- and
- wherein the inter-class rule comprises an at least one of (i) a versioning-pass rule, and (ii) a-versioning-match rule;
- and
- based on a result of the applying step, for each respective classified section, performing an at least one of (i) versioning the classified section;
- (ii) adjusting a data value of a data field in the classified section;
- and (iii) adjusting a confidence value of a data field of the classified section.
9 more elements
An automated document-analysis method comprising the steps of:
- receiving an electronic document comprising a plurality of sections;
- receiving a corresponding pre-classification value for each section of the plurality of sections;
- classifying the electronic document, such that a corresponding classification value is determined for each section of the plurality of sections;
- comparing the corresponding pre-qualification value for at least one section of the plurality of sections against the determined corresponding classification value for the section;
- if the corresponding pre-qualification value does not match the determined corresponding classification value, performing an at least one of (i) identifying the section as mis-classified;
- (ii) designating the section for a manual review;
- and (iii) designating the section for an automated review.
4 more elements
An automated document-analysis method comprising the steps of:
- receiving an electronic document comprising a plurality of sections;
- classifying the electronic document, such that a corresponding classification value is determined for each section of the plurality of sections;
- extracting an at least one attribute of each section of the plurality of sections;
- comparing the at least one extracted attribute of an at least one section of the plurality of sections against an expected attribute determined based on the corresponding classification value for the section;
- if the corresponding at least one extracted attribute does not match the expected attribute determined based on the corresponding classification value, performing an at least one of (i) identifying the section as mis-classified;
- (ii) designating the section for a manual review;
- and (iii) designating the section for an automated review
4 more elements
An automated document-analysis method comprising the steps of:
- receiving an electronic document comprising a plurality of sections;
- classifying the electronic document using a first classifier, such that a corresponding first classification value is determined for each section of the plurality of sections;
- classifying the electronic document using a second classifier, such that a corresponding second classification value is determined for each section of the plurality of sections;
- comparing the two determined classification values for at least one section of the plurality of sections;
- and
- if the two determined classification values do not match each other, performing an at least one of (i) identifying which of the two determined classification values is correct;
- (ii) identifying the section as mis-classified;
- (iii) designating the section for a manual review;
- and (iv) designating the section for an automated review.
6 more elements
Dependencies are inferred from the claim text.