Secure Document Data Capture System with Redacted Exception Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document data capture systems face errors in character recognition, especially when dealing with low-contrast or distorted document images, and require extensive security measures to protect confidential information during exception handling, which is resource-intensive.
Innovation Solution
A secure document data capture system that generates exception images with redacted context portions, allowing user input for suspect characters, and only substitutes replacements when confirmed by multiple independent clients, ensuring accuracy and security without excessive resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional exception handling systems display the full document image to human operators for correcting mis-recognized characters, then character recognition accuracy is improved, but security resources and system complexity increase significantly to protect confidential information
Solution Approach 1:
The patent segments the document image into two distinct portions: an exception portion containing only the suspect characters that need correction, and a redacted portion containing confidential information that is replaced with redaction marks. This segmentation allows human operators to correct character recognition errors while confidential information remains protected, resolving the contradiction between improving recognition accuracy and reducing security system complexity.
Solution Approach 2:
The patent extracts only the necessary exception data (suspect characters and their context) from the full document image, separating it from confidential information. By taking out only the minimal required information for exception handling and redacting the rest, the system achieves accurate character correction without exposing confidential data, thereby reducing the complexity of security measures needed.
2Measurement precision
If traditional exception handling systems transmit the full document image to human operators, then accurate character correction is enabled, but extensive security measures and resources are required to protect confidential information
Solution Approach 1:
The system segments the document transmission by sending only the exception portion (containing suspect characters) and a redacted version of the confidential portion to human operators. This segmentation reduces the amount of data that requires security protection during transmission, enabling accurate character correction while minimizing security resource consumption.
Solution Approach 2:
The system extracts and transmits only the essential exception information needed for character correction, removing confidential information from the transmitted data stream. This extraction approach maintains character correction accuracy while reducing the security resources needed to protect transmitted data.
3Measurement precision
If the system displays the full document image for exception handling, then human operators can accurately identify mis-recognized characters, but the system requires extensive security policies and procedures to protect confidential information
Solution Approach 1:
The patent implements segmentation by displaying separate portions of the document to human operators: the exception portion showing suspect characters for identification and correction, and a redacted portion where confidential information is masked. This segmentation allows operators to accurately identify mis-recognized characters without being exposed to confidential information, thereby reducing the complexity of security policies and procedures required.
Data Source
AI summary
An invoice processing system generates a data structure, comprising a data field value associated with each of a plurality of identified data fields, from a document image comprising text representing each of such data field values. The secure document data capture system includes a character recognition system receiving the document image and recognizing characters within the text to generate, for each of the identified data fields, a data field value for association therewith. A validation engine identifies a subset of the identified data fields which can be referred to as exception data fields due to failure to comply with a validation rule. An exception handling system provides an exception image to a first client system. The exception image comprising a portion of the document image which includes text of the at least one suspect character within the exception data field with a context portion of the document image redacted. The context portion of the document image is a portion of the document image which comprises text which discloses a meaning of the at least one suspect character or the data field value. The exception handling system receives, from the first client system, user input of a replacement character for at least each suspect character. The secure data capture system then generates the data structure utilizing, for each identified data field, the data field values generated by the character recognition system with substitution of the replacement character from the exception handling system for each suspect character.