Character Recognition Strikethrough Noise Rejection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data input systems incorrectly identify scan noise as strikethrough, leading to rejection of accurate character recognition results, which can result in incorrect cancellation of character strings marked for deletion.
Innovation Solution
An information processing apparatus that includes a character recognition section, a strikethrough detection section, a matching section, and a control section to verify the accuracy of character recognition results by matching them with human input, even when a strikethrough is detected, ensuring correct character recognition results are not rejected unnecessarily.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If scan noise is detected as strikethrough, then the character recognition result is rejected, but the correct character recognition result is lost
Solution Approach 1:
The system performs matching between character recognition results and human inputs to obtain feedback on the accuracy of recognition. When a strikethrough is detected, the system uses this feedback mechanism to verify whether the character recognition result should be rejected, preventing erroneous rejection due to noise detection.
Solution Approach 2:
The matching section acts as an intermediary between the strikethrough detection and the final rejection decision. It compares character recognition results with human inputs to determine whether the detected strikethrough is genuine or noise, thereby mediating the rejection process to avoid losing correct recognition results.
2Measurement precision
If character recognition result is rejected when strikethrough is detected, then false positive noise is rejected, but actual strikethrough cases are also rejected
Solution Approach 1:
The matching process provides feedback by comparing automated character recognition with human inputs. This feedback allows the system to distinguish between genuine strikethroughs (where human input would differ from recognition) and noise (where human input matches recognition), improving both precision and reliability.
Solution Approach 2:
The system changes the decision parameter from a simple binary rejection based on strikethrough detection to a comparative parameter based on matching results. By evaluating whether recognition results match human inputs, the system adjusts its decision criteria to improve both precision and reliability simultaneously.
3Reliability
If multiple human verifications are required, then accuracy is improved, but processing time and cost increase
Solution Approach 1:
The system extracts and utilizes existing human inputs (such as manual corrections or annotations) as verification data without requiring additional human verification steps. By taking out and reusing available human input data for matching, the system maintains accuracy while avoiding the time and cost of multiple verifications.
Solution Approach 2:
The system performs self-verification by matching its own character recognition results against available human inputs. This self-service mechanism allows the system to validate its results without requiring external human verification, thereby maintaining reliability while reducing processing time and cost.
Data Source
AI summary
An information processing apparatus includes a character recognition section that performs character recognition of an input image to output a character recognition result, a receiving section that receives an input of a character recognition result by a person on the input image, a detection section that detects a strikethrough from the input image, a matching section that matches the character recognition result output by the character recognition section with the character recognition result by the person, which is received by the receiving section, and a control section that performs control for causing the matching section to perform matching so as to obtain a final character recognition result based on a result of the matching, in a case where the detection section detects the strikethrough.