Document Attribute Extraction Using Dynamic Template Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing paper-based documents in electronic applications face inefficiencies due to the need for human confirmation, low extraction precision of attributes, and potential work-related losses from erroneous automatic identification, especially when templates are not optimally selected or changed.
Innovation Solution
A computer system that extracts attributes from document images using a processor and storage device with predefined templates, calculates credibility evaluation values, and adjusts confirmation operations based on these values to ensure high precision and minimize errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic identification is performed using a device, then work efficiency is improved and cost is reduced, but work-related losses may occur due to erroneous automatic identification
Solution Approach 1:
The system calculates an evaluation value indicating credibility of extracted attributes and compares it with a threshold to determine the type of confirmation operation. This feedback mechanism allows the system to automatically adjust the level of human review based on confidence levels, maintaining high productivity while minimizing errors through selective human verification.
Solution Approach 2:
The confirmation operation type is dynamically adjusted based on the evaluation value. When the evaluation value exceeds the threshold, automated processing is used; when it falls below, human confirmation is triggered. This dynamic adaptation resolves the contradiction by allowing the system to switch between automated high-speed processing and human-verified high-reliability processing.
2Ease of manufacture
If templates are preset for each type of bill, then attribute extraction can be performed, but extraction precision is low when templates are not optimally selected or when bill formats change
Solution Approach 1:
The system changes the parameter of template selection from static presetting to dynamic selection based on document characteristics. By analyzing the document image and automatically selecting or generating appropriate templates, the system maintains high extraction precision even when bill formats change, without requiring manual template updates for each format variation.
Solution Approach 2:
The template management system is designed to be universal, handling multiple bill types and format variations through a single automated template selection mechanism. This multi-functional approach eliminates the need for separate manual template preparation for each bill type while maintaining high extraction precision across diverse document formats.
3Reliability
If human staff confirm attributes on paper-based documents, then work precision is maintained, but work efficiency is low and cost is high
Solution Approach 1:
Instead of requiring full human confirmation for all documents, the system applies partial human action only when necessary. By calculating evaluation values and comparing them with thresholds, the system identifies cases where human confirmation is needed (low confidence cases) and processes them accordingly, while automatically handling high-confidence cases. This partial automation approach maintains high accuracy while significantly improving工作效率.
4Productivity
If OCR technology is used to read attributes from paper-based documents, then manual work is reduced, but extraction precision varies when templates are not optimally selected
Solution Approach 1:
The system performs preliminary actions by calculating evaluation values for extracted attributes before final confirmation. This preliminary assessment allows the system to identify low-confidence extractions that require human review, ensuring high precision while maintaining automation benefits. The evaluation calculation acts as a preliminary filter that prepares the data for either automated acceptance or human verification.
Data Source
AI summary
A computer that extracts an attribute which is a text string contained in an predetermined examination target document stores template information for managing a plurality of templates in which a type of attribute is defined, executes a text recognition process on image data of the document, extracts an attribute corresponding to the type of attribute using a result of the text recognition process and the plurality of templates, selects a template based on the extracted attribute, generates output information that includes the attribute extracted using the selected template and is used for the examination; determines a type of confirmation operation performed on the output information, before the examination, based on a comparison result between an evaluation value indicating credibility of the output information and a threshold, and corrects the determined type of confirmation operation based on the text string contained in the document.