Digitized Text Certainty Scoring for OCR Reliability Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition systems struggle to accurately gauge the reliability of extracted data due to arbitrarily defined confidence score thresholds, leading to high false positive and false negative rates.
Innovation Solution
A certainty algorithm that determines reliability indicators by using content and positional confidence scores to generate certainty scores and quotients, dynamically assessing data reliability through contextual information and normalization, reducing false positives and negatives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If arbitrarily defined confidence score thresholds are used in existing text recognition systems, then the system can provide a measure of confidence in its output, but the thresholds are too high or too low to prevent significant numbers of false positives and false negatives, resulting in inability to accurately gauge reliability of extracted data
Solution Approach 1:
The patent transforms the single confidence score parameter into multiple certainty score parameters (content certainty, positional certainty, span certainty) that are calculated using different formulas and thresholds. This parameter transformation allows the system to assess reliability from multiple dimensions rather than relying on a single imprecise threshold, thereby resolving the contradiction between reliability assessment and measurement precision
Solution Approach 2:
The patent segments the monolithic confidence score into distinct certainty components: content certainty (word-level accuracy), positional certainty (field location accuracy), and span certainty (text boundary accuracy). Each segment has its own calculation method and threshold, enabling precise measurement in each dimension while collectively providing comprehensive reliability assessment
2Reliability
If existing text recognition systems use single confidence scores, then the system can provide a simplified reliability indicator, but the single score cannot accurately distinguish between different types of recognition errors, leading to high false positive and false negative rates
Solution Approach 1:
The patent divides the error detection task into three specialized certainty algorithms, each focusing on a specific aspect: content certainty for word recognition errors, positional certainty for field location errors, and span certainty for boundary errors. This segmentation allows each algorithm to be optimized for its specific purpose, improving error detection accuracy while keeping individual algorithm complexity manageable
Solution Approach 2:
The patent creates a universal certainty scoring framework that can handle multiple types of recognition errors through a unified multi-dimensional approach. The framework processes different error types (content, position, span) using consistent methodology while maintaining the ability to distinguish between them, achieving both accuracy and systematic complexity management
Data Source
AI summary
Techniques for improving the reliability of digitized text may comprise receiving data corresponding to a data extraction program. The techniques may further comprise executing a confidence algorithm to generate a dictionary of words from fields in a file, determine, based at least in part on a confidence score from the dictionary of words, a content confidence score corresponding to a first field of the file, and determine, based on the content confidence score and a position confidence score, a certainty score for the first field. The techniques may further comprise determining, based at least in part on the certainty score, a certainty quotient for the file indicating a reliability of the data corresponding to the data extraction program, and generating a data object that indicates the certainty quotient. These techniques generate reliable indications of the accuracy of digitized text extracted from a file and substantially reduce false positives/negatives.