OCR Correction Learning for Flexible Scan Filename Setting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing systems for setting filenames based on character recognition results of scanned forms lack flexibility in allowing users to dynamically set conditions, and existing OCR technologies do not effectively handle errors or allow for user corrections.

Innovation Solution

An information processing apparatus and method that includes a multifunction peripheral (MFP) and a client personal computer (PC) connected via a network, utilizing a cloud-based service for image analysis, allowing users to correct and learn from OCR results to dynamically set filenames and metadata, with the system learning from user corrections to improve future recognition accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If pre-defined conditions are set for prohibited characters in filenames, then filename management reliability is improved, but user flexibility and ease of operation deteriorate

Engineering Contradiction:
Improvefilename management reliabilityVSAvoiduser flexibility
Core Design Contradiction:
ReliabilityVSEase of operation

Solution Approach 1:

The system dynamically adjusts filename conditions based on the actual recognition results. Instead of using fixed pre-defined prohibited character lists, the system automatically determines appropriate conditions by analyzing the OCR output, allowing the filename rules to adapt flexibly to different recognition scenarios while maintaining reliable filename management

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes the parameters of filename conditions based on recognition results. By modifying the prohibited characters and filename rules dynamically according to the actual OCR output, the system achieves both reliability in filename management and flexibility in adapting to different form images

Inventive Principle:
Principle #35Parameter changes

2Stability of the object's composition

If strict filename rules are imposed in advance, then filename consistency is improved, but adaptability to different form images deteriorates

Engineering Contradiction:
Improvefilename consistencyVSAvoidadaptability to different form images
Core Design Contradiction:
Stability of the object's compositionVSAdaptability or versatility

Solution Approach 1:

The system employs dynamic filename condition determination that adjusts to different form image types. By automatically analyzing recognition results and setting appropriate filename rules on-the-fly, the system maintains consistent naming conventions for each form type while adapting to the diversity of different forms

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system performs self-service by automatically determining appropriate filename conditions based on the recognition results without requiring manual pre-configuration. The system serves itself by learning from the OCR output and autonomously setting suitable filename rules, achieving both consistency and adaptability

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3890295B1Information processing apparatus for obtaining character string
Publication Date: 2026.05.06 CANON KK
  • EP3890295B1 patent drawingFigure 1
  • EP3890295B1 patent drawingFigure 2
  • EP3890295B1 patent drawingFigure 3

AI summary

An information processing apparatus comprising: character recognition means for obtaining a character recognition result by performing character recognition processing on a text region in a first scan image; and learning means for, if a correction is made to at least a part of a character string of the character recognition result in setting attribute information about the first scan image by using the character recognition result obtained by the character recognition means, learning correction content of the correction, wherein the character recognition means is configured to, if the character recognition processing is performed on a text region in a second scan image, correct a character recognition result of the text region in the second scan image based on the correction content learned by the learning means, and output the corrected character recognition result.