Text Recognition Confidence Matrix Regular Expression Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for text recognition, especially from historical manuscripts and spoken text, are inefficient and require manual transcription, struggling with variations in writing styles and accents, and are not adaptable for complex search patterns.
Innovation Solution
A method using a confidence matrix and regular expressions to directly decode symbol sequences, allowing for fast and fault-tolerant text recognition without manual transcription, enabling complex search patterns and parsing of structured and semi-structured data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual transcription is used to convert text to computer code, then text recognition accuracy is improved, but time consumption and cost increase significantly
Solution Approach 1:
The patent replaces the mechanical manual transcription process with an automated optical character recognition (OCR) system that uses image processing algorithms to directly convert images of text into computer code, eliminating the need for human transcribers while maintaining high accuracy through sophisticated pattern recognition
Solution Approach 2:
The patent introduces an intermediate processing stage that converts text images into structured data formats with confidence scores, serving as a bridge between image recognition and final text output, allowing for automated quality assessment and correction without manual intervention
2Measurement precision
If direct search based on sample sequences is used, then text recognition is performed, but processing speed becomes very slow
Solution Approach 1:
The patent performs preliminary processing by converting text images into structured confidence score data before the actual recognition and search operations, organizing the data in advance to enable faster querying and comparison during the recognition process
Solution Approach 2:
The patent replaces slow sequential sample-by-sample comparison with optimized algorithms that process confidence score matrices efficiently, using mathematical operations to rapidly identify matching text patterns without exhaustive searching
3Reliability
If manual transcription is performed to achieve text recognition, then text can be searched, but the process is error-prone and resource-intensive
Solution Approach 1:
The patent implements self-service through automated confidence score calculation and quality assessment mechanisms that evaluate and validate recognized text without human intervention, automatically identifying and correcting errors while maintaining high reliability
Solution Approach 2:
The patent incorporates feedback loops where confidence scores from the recognition process are used to guide further processing decisions, allowing the system to self-correct and validate results, reducing errors while simplifying the overall process through automated decision-making
Data Source
AI summary
The invention refers to a method for text recognition, wherein the method is executed by a processor of a computing device and comprises steps of providing a confidence matrix, wherein the confidence matrix is a digital representation of an input sequence, entering a regular expression, searching for a symbol sequence of the input sequence that matches the regular expression, wherein a score value is computed by the processor using confidence values of the confidence matrix, wherein the score value is an indication of the quality of the matching between the symbol sequence of the input sequence and the regular expression. Further, the invention relates to a computer program product which when executed by a processor of a computing device performs the method.