Pattern-Based OCR Correction for Unique Character Strings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCR methods face difficulties in accurately processing character strings like serial numbers, member IDs, and patient IDs in reports, as these unique strings are hard to register in word lists and often result in low accuracy due to their distinct nature.
Innovation Solution
The system determines if a character string from the OCR process matches a word dictionary and, if not, applies a pattern-based adjustment to the recognition process by detecting and weighting character string patterns, allowing for improved accuracy in processing non-standard character strings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a word list matching method is used for OCR correction, then common words can be corrected effectively, but unique character strings like serial numbers cannot be processed accurately
Solution Approach 1:
The patent segments the character string verification process into two distinct phases: first checking against a word list for common words, then applying pattern matching for unique strings. This segmentation allows each method to be optimized for its appropriate use case, resolving the contradiction between handling common words and unique strings effectively
Solution Approach 2:
The patent introduces pattern templates as an intermediary mechanism between the word list and unique character strings. These patterns serve as a bridge that can match both conventional words and unique identifiers like serial numbers, enabling the system to handle both types of character strings with a unified approach
2Measurement precision
If OCR processing is performed on all character strings, then processing time increases, but accuracy for unique strings improves
Solution Approach 1:
The patent applies preliminary action by first checking character strings against a word list before performing full OCR processing. This preliminary check quickly identifies common words that can be corrected through simple matching, avoiding the need for time-consuming OCR processing on every character string while still maintaining accuracy for unique strings that require full processing
Data Source
AI summary
The present disclosure determines whether or not a character string of a result obtained by a character recognition process matches a word of a word dictionary; and when a pattern that is similar to a predefined arrangement pattern of a character type is present in the character string of the result obtained by the character recognition process that is determined not to match a word of the word dictionary, changes the character recognition process for the character string based on the pattern.