Document Division Accuracy via Machine Learning and User Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for dividing scan images into documents often result in incorrect division positions, leading to incomplete document files and requiring users to manually correct and merge files, causing inconvenience.
Innovation Solution
An information processing apparatus that uses a division unit to determine division positions based on text data extraction and machine learning classifiers to correctly separate scan images into document files, and provides a correction mechanism for users to adjust division positions through a graphical user interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a division position is determined based on analysis results, then the scan images can be divided into document files, but the division position may be determined incorrectly leading to incomplete document files
Solution Approach 1:
The system provides feedback to users about determined division positions and allows users to provide correction inputs when division positions are incorrect. This feedback loop enables the system to learn from user corrections and improve future division accuracy while maintaining automated operation.
Solution Approach 2:
The system performs preliminary division based on analysis results before user review, allowing users to correct only specific incorrect divisions rather than manually dividing entire documents. This preliminary action maintains productivity while enabling reliability improvement through user feedback.
2Measurement precision
If incorrect division positions are determined, then document files become incomplete, but users must manually correct and merge files causing inconvenience
Solution Approach 1:
The system extracts and presents only the specific division positions that require user correction, rather than requiring users to review and correct entire documents. This extraction approach maintains measurement precision while significantly reducing user effort by focusing corrections only where needed.
3Reliability
If users need to find and merge document files to correct division errors, then the process becomes complex, but accurate division is achieved
Solution Approach 1:
The system combines the division determination function with a user feedback mechanism into a single integrated process. Instead of requiring separate operations to find files and merge them, the system handles the entire correction process through a unified interface that maintains reliability while reducing complexity.
Data Source
AI summary
An information processing apparatus obtains scan images by collectively scanning a plurality of documents, each including page(s), divides the scan images into divided scan images corresponding to respective documents, converts the divided scan images into formatted files, thereby generating a plurality of files, saving the files individually, manages a common file identifier, ordinal position information of files, and identifying information of the individually saved files in association with each other, and displays, if a user instructs correcting one of the individually saved files, thumbnails of page images corresponding to the individually saved plurality of documents identified by the information managed in association with the same identifier as an identifier associated with information identifying the one file to be corrected. The thumbnails are displayed, based on the ordinal position information of the plurality of files, in an order in which the pages forming the plurality of files were scanned.