OCR Attribute Association for Variable Document Layouts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) methods fail to provide supplementary information when character strings that supplement the content for an attribute are positioned differently than expected, making it difficult to utilize the extracted attribute values effectively in subsequent processes.
Innovation Solution
An information processing apparatus that extracts attribute values from OCR results and associates them with sub-attributes using predefined extraction and association rules, even when the sub-attribute values are not initially acquired together with the main attribute values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If attribute values are extracted using fixed extraction rules, then extraction efficiency is improved, but completeness of information is worsened when sub-attributes appear at different positions
Solution Approach 1:
The system performs preliminary extraction of attribute values according to fixed rules, then subsequently associates sub-attributes with main attributes through association rules. This two-stage approach maintains extraction efficiency while capturing complete information even when sub-attributes appear at different positions than expected.
Solution Approach 2:
Association rules act as intermediaries between the extracted attribute values and the required complete information. These rules enable the system to link sub-attributes to main attributes, filling in missing information without requiring changes to the initial extraction process.
2Measurement precision
If strict position-based extraction is used, then extraction accuracy is improved, but adaptability to different document formats is worsened
Solution Approach 1:
The system transitions from static position-based extraction to a dynamic two-stage process where extraction rules provide initial accurate extraction, and association rules dynamically adapt to different document formats by linking attributes based on their semantic relationships rather than fixed positions.
Solution Approach 2:
The system changes the parameter of extraction from position-dependent to relationship-dependent. By using association rules that connect attributes based on their semantic meaning rather than fixed positions, the system maintains accuracy while adapting to various document formats.
Data Source
AI summary
An information processing apparatus includes a processor configured to: extract attribute values for plural attributes from an optical character recognition (OCR) result for an image in accordance with an extraction rule determined in advance, and generate an extraction result in which the extracted attribute values and the respective attributes are correlated with each other; and perform control so as to associate, in accordance with an association rule for associating an attribute value for a sub attribute with a different attribute, the attribute value for the sub attribute with the different attribute included in the extraction result, the sub attribute being an attribute for which an attribute value is extracted from the image and which supplements a content for the different attribute.